qemu/usb-linux.c

1905 lines
52 KiB
C
Raw Normal View History

/*
* Linux host USB redirector
*
* Copyright (c) 2005 Fabrice Bellard
*
* Copyright (c) 2008 Max Krasnyansky
* Support for host device auto connect & disconnect
* Major rewrite to support fully async operation
*
* Copyright 2008 TJ <linux@tjworld.net>
* Added flexible support for /dev/bus/usb /sys/bus/usb/devices in addition
* to the legacy /proc/bus/usb USB device discovery and handling
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "qemu-common.h"
#include "qemu-timer.h"
#include "monitor.h"
#include "sysemu.h"
#include "trace.h"
#include <dirent.h>
#include <sys/ioctl.h>
#include <linux/usbdevice_fs.h>
#include <linux/version.h>
#include "hw/usb.h"
/* We redefine it to avoid version problems */
struct usb_ctrltransfer {
uint8_t bRequestType;
uint8_t bRequest;
uint16_t wValue;
uint16_t wIndex;
uint16_t wLength;
uint32_t timeout;
void *data;
};
typedef int USBScanFunc(void *opaque, int bus_num, int addr, const char *port,
int class_id, int vendor_id, int product_id,
const char *product_name, int speed);
//#define DEBUG
#ifdef DEBUG
#define DPRINTF printf
#else
#define DPRINTF(...)
#endif
#define PRODUCT_NAME_SZ 32
#define MAX_PORTLEN 16
/* endpoint association data */
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
#define ISO_FRAME_DESC_PER_URB 32
/* devio.c limits single requests to 16k */
#define MAX_USBFS_BUFFER_SIZE 16384
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
typedef struct AsyncURB AsyncURB;
struct endp_data {
uint8_t halted;
uint8_t iso_started;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
AsyncURB *iso_urb;
int iso_urb_idx;
int iso_buffer_used;
int inflight;
};
struct USBAutoFilter {
uint32_t bus_num;
uint32_t addr;
char *port;
uint32_t vendor_id;
uint32_t product_id;
};
typedef struct USBHostDevice {
USBDevice dev;
int fd;
int hub_fd;
int hub_port;
uint8_t descr[8192];
int descr_len;
int closing;
uint32_t iso_urb_count;
Notifier exit;
struct endp_data ep_in[USB_MAX_ENDPOINTS];
struct endp_data ep_out[USB_MAX_ENDPOINTS];
QLIST_HEAD(, AsyncURB) aurbs;
/* Host side address */
int bus_num;
int addr;
char port[MAX_PORTLEN];
struct USBAutoFilter match;
int seen, errcount;
QTAILQ_ENTRY(USBHostDevice) next;
} USBHostDevice;
static QTAILQ_HEAD(, USBHostDevice) hostdevs = QTAILQ_HEAD_INITIALIZER(hostdevs);
static int usb_host_close(USBHostDevice *dev);
static int parse_filter(const char *spec, struct USBAutoFilter *f);
static void usb_host_auto_check(void *unused);
static int usb_host_read_file(char *line, size_t line_size,
const char *device_file, const char *device_name);
static int usb_linux_update_endp_table(USBHostDevice *s);
static int usb_host_usbfs_type(USBHostDevice *s, USBPacket *p)
{
static const int usbfs[] = {
[USB_ENDPOINT_XFER_CONTROL] = USBDEVFS_URB_TYPE_CONTROL,
[USB_ENDPOINT_XFER_ISOC] = USBDEVFS_URB_TYPE_ISO,
[USB_ENDPOINT_XFER_BULK] = USBDEVFS_URB_TYPE_BULK,
[USB_ENDPOINT_XFER_INT] = USBDEVFS_URB_TYPE_INTERRUPT,
};
uint8_t type = p->ep->type;
assert(type < ARRAY_SIZE(usbfs));
return usbfs[type];
}
static int usb_host_do_reset(USBHostDevice *dev)
{
struct timeval s, e;
uint32_t usecs;
int ret;
gettimeofday(&s, NULL);
ret = ioctl(dev->fd, USBDEVFS_RESET);
gettimeofday(&e, NULL);
usecs = (e.tv_sec - s.tv_sec) * 1000000;
usecs += e.tv_usec - s.tv_usec;
if (usecs > 1000000) {
/* more than a second, something is fishy, broken usb device? */
fprintf(stderr, "husb: device %d:%d reset took %d.%06d seconds\n",
dev->bus_num, dev->addr, usecs / 1000000, usecs % 1000000);
}
return ret;
}
static struct endp_data *get_endp(USBHostDevice *s, int pid, int ep)
{
struct endp_data *eps = pid == USB_TOKEN_IN ? s->ep_in : s->ep_out;
assert(pid == USB_TOKEN_IN || pid == USB_TOKEN_OUT);
assert(ep > 0 && ep <= USB_MAX_ENDPOINTS);
return eps + ep - 1;
}
static int is_isoc(USBHostDevice *s, int pid, int ep)
{
return usb_ep_get_type(&s->dev, pid, ep) == USB_ENDPOINT_XFER_ISOC;
}
static int is_valid(USBHostDevice *s, int pid, int ep)
{
return usb_ep_get_type(&s->dev, pid, ep) != USB_ENDPOINT_XFER_INVALID;
}
static int is_halted(USBHostDevice *s, int pid, int ep)
{
return get_endp(s, pid, ep)->halted;
}
static void clear_halt(USBHostDevice *s, int pid, int ep)
{
trace_usb_host_ep_clear_halt(s->bus_num, s->addr, ep);
get_endp(s, pid, ep)->halted = 0;
}
static void set_halt(USBHostDevice *s, int pid, int ep)
{
if (ep != 0) {
trace_usb_host_ep_set_halt(s->bus_num, s->addr, ep);
get_endp(s, pid, ep)->halted = 1;
}
}
static int is_iso_started(USBHostDevice *s, int pid, int ep)
{
return get_endp(s, pid, ep)->iso_started;
}
static void clear_iso_started(USBHostDevice *s, int pid, int ep)
{
trace_usb_host_ep_stop_iso(s->bus_num, s->addr, ep);
get_endp(s, pid, ep)->iso_started = 0;
}
static void set_iso_started(USBHostDevice *s, int pid, int ep)
{
struct endp_data *e = get_endp(s, pid, ep);
trace_usb_host_ep_start_iso(s->bus_num, s->addr, ep);
if (!e->iso_started) {
e->iso_started = 1;
e->inflight = 0;
}
}
static int change_iso_inflight(USBHostDevice *s, int pid, int ep, int value)
{
struct endp_data *e = get_endp(s, pid, ep);
e->inflight += value;
return e->inflight;
}
static void set_iso_urb(USBHostDevice *s, int pid, int ep, AsyncURB *iso_urb)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
get_endp(s, pid, ep)->iso_urb = iso_urb;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
static AsyncURB *get_iso_urb(USBHostDevice *s, int pid, int ep)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
return get_endp(s, pid, ep)->iso_urb;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
static void set_iso_urb_idx(USBHostDevice *s, int pid, int ep, int i)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
get_endp(s, pid, ep)->iso_urb_idx = i;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
static int get_iso_urb_idx(USBHostDevice *s, int pid, int ep)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
return get_endp(s, pid, ep)->iso_urb_idx;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
static void set_iso_buffer_used(USBHostDevice *s, int pid, int ep, int i)
{
get_endp(s, pid, ep)->iso_buffer_used = i;
}
static int get_iso_buffer_used(USBHostDevice *s, int pid, int ep)
{
return get_endp(s, pid, ep)->iso_buffer_used;
}
/*
* Async URB state.
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
* We always allocate iso packet descriptors even for bulk transfers
* to simplify allocation and casts.
*/
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
struct AsyncURB
{
struct usbdevfs_urb urb;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
struct usbdevfs_iso_packet_desc isocpd[ISO_FRAME_DESC_PER_URB];
USBHostDevice *hdev;
QLIST_ENTRY(AsyncURB) next;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
/* For regular async urbs */
USBPacket *packet;
int more; /* large transfer, more urbs follow */
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
/* For buffered iso handling */
int iso_frame_idx; /* -1 means in flight */
};
static AsyncURB *async_alloc(USBHostDevice *s)
{
AsyncURB *aurb = g_malloc0(sizeof(AsyncURB));
aurb->hdev = s;
QLIST_INSERT_HEAD(&s->aurbs, aurb, next);
return aurb;
}
static void async_free(AsyncURB *aurb)
{
QLIST_REMOVE(aurb, next);
g_free(aurb);
}
static void do_disconnect(USBHostDevice *s)
{
usb_host_close(s);
usb_host_auto_check(NULL);
}
static void async_complete(void *opaque)
{
USBHostDevice *s = opaque;
AsyncURB *aurb;
int urbs = 0;
while (1) {
USBPacket *p;
int r = ioctl(s->fd, USBDEVFS_REAPURBNDELAY, &aurb);
if (r < 0) {
if (errno == EAGAIN) {
if (urbs > 2) {
fprintf(stderr, "husb: %d iso urbs finished at once\n", urbs);
}
return;
}
if (errno == ENODEV) {
if (!s->closing) {
trace_usb_host_disconnect(s->bus_num, s->addr);
do_disconnect(s);
}
return;
}
perror("USBDEVFS_REAPURBNDELAY");
return;
}
DPRINTF("husb: async completed. aurb %p status %d alen %d\n",
aurb, aurb->urb.status, aurb->urb.actual_length);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
/* If this is a buffered iso urb mark it as complete and don't do
anything else (it is handled further in usb_host_handle_iso_data) */
if (aurb->iso_frame_idx == -1) {
int inflight;
int pid = (aurb->urb.endpoint & USB_DIR_IN) ?
USB_TOKEN_IN : USB_TOKEN_OUT;
int ep = aurb->urb.endpoint & 0xf;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
if (aurb->urb.status == -EPIPE) {
set_halt(s, pid, ep);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
aurb->iso_frame_idx = 0;
urbs++;
inflight = change_iso_inflight(s, pid, ep, -1);
if (inflight == 0 && is_iso_started(s, pid, ep)) {
fprintf(stderr, "husb: out of buffers for iso stream\n");
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
continue;
}
p = aurb->packet;
trace_usb_host_urb_complete(s->bus_num, s->addr, aurb, aurb->urb.status,
aurb->urb.actual_length, aurb->more);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
if (p) {
switch (aurb->urb.status) {
case 0:
p->result += aurb->urb.actual_length;
break;
case -EPIPE:
set_halt(s, p->pid, p->ep->nr);
p->result = USB_RET_STALL;
break;
default:
p->result = USB_RET_NAK;
break;
}
if (aurb->urb.type == USBDEVFS_URB_TYPE_CONTROL) {
trace_usb_host_req_complete(s->bus_num, s->addr, p->result);
usb_generic_async_ctrl_complete(&s->dev, p);
} else if (!aurb->more) {
trace_usb_host_req_complete(s->bus_num, s->addr, p->result);
usb_packet_complete(&s->dev, p);
}
}
async_free(aurb);
}
}
static void usb_host_async_cancel(USBDevice *dev, USBPacket *p)
{
USBHostDevice *s = DO_UPCAST(USBHostDevice, dev, dev);
AsyncURB *aurb;
QLIST_FOREACH(aurb, &s->aurbs, next) {
if (p != aurb->packet) {
continue;
}
DPRINTF("husb: async cancel: packet %p, aurb %p\n", p, aurb);
/* Mark it as dead (see async_complete above) */
aurb->packet = NULL;
int r = ioctl(s->fd, USBDEVFS_DISCARDURB, aurb);
if (r < 0) {
DPRINTF("husb: async. discard urb failed errno %d\n", errno);
}
}
}
static int usb_host_open_device(int bus, int addr)
{
const char *usbfs = NULL;
char filename[32];
struct stat st;
int fd, rc;
rc = stat("/dev/bus/usb", &st);
if (rc == 0 && S_ISDIR(st.st_mode)) {
/* udev-created device nodes available */
usbfs = "/dev/bus/usb";
} else {
/* fallback: usbfs mounted below /proc */
usbfs = "/proc/bus/usb";
}
snprintf(filename, sizeof(filename), "%s/%03d/%03d",
usbfs, bus, addr);
fd = open(filename, O_RDWR | O_NONBLOCK);
if (fd < 0) {
fprintf(stderr, "husb: open %s: %s\n", filename, strerror(errno));
}
return fd;
}
static int usb_host_claim_port(USBHostDevice *s)
{
#ifdef USBDEVFS_CLAIM_PORT
char *h, hub_name[64], line[1024];
int hub_addr, ret;
snprintf(hub_name, sizeof(hub_name), "%d-%s",
s->match.bus_num, s->match.port);
/* try strip off last ".$portnr" to get hub */
h = strrchr(hub_name, '.');
if (h != NULL) {
s->hub_port = atoi(h+1);
*h = '\0';
} else {
/* no dot in there -> it is the root hub */
snprintf(hub_name, sizeof(hub_name), "usb%d",
s->match.bus_num);
s->hub_port = atoi(s->match.port);
}
if (!usb_host_read_file(line, sizeof(line), "devnum",
hub_name)) {
return -1;
}
if (sscanf(line, "%d", &hub_addr) != 1) {
return -1;
}
s->hub_fd = usb_host_open_device(s->match.bus_num, hub_addr);
if (s->hub_fd < 0) {
return -1;
}
ret = ioctl(s->hub_fd, USBDEVFS_CLAIM_PORT, &s->hub_port);
if (ret < 0) {
close(s->hub_fd);
s->hub_fd = -1;
return -1;
}
trace_usb_host_claim_port(s->match.bus_num, hub_addr, s->hub_port);
return 0;
#else
return -1;
#endif
}
static void usb_host_release_port(USBHostDevice *s)
{
if (s->hub_fd == -1) {
return;
}
#ifdef USBDEVFS_RELEASE_PORT
ioctl(s->hub_fd, USBDEVFS_RELEASE_PORT, &s->hub_port);
#endif
close(s->hub_fd);
s->hub_fd = -1;
}
static int usb_host_disconnect_ifaces(USBHostDevice *dev, int nb_interfaces)
{
/* earlier Linux 2.4 do not support that */
#ifdef USBDEVFS_DISCONNECT
struct usbdevfs_ioctl ctrl;
int ret, interface;
for (interface = 0; interface < nb_interfaces; interface++) {
ctrl.ioctl_code = USBDEVFS_DISCONNECT;
ctrl.ifno = interface;
ctrl.data = 0;
ret = ioctl(dev->fd, USBDEVFS_IOCTL, &ctrl);
if (ret < 0 && errno != ENODATA) {
perror("USBDEVFS_DISCONNECT");
return -1;
}
}
#endif
return 0;
}
static int usb_linux_get_num_interfaces(USBHostDevice *s)
{
char device_name[64], line[1024];
int num_interfaces = 0;
sprintf(device_name, "%d-%s", s->bus_num, s->port);
if (!usb_host_read_file(line, sizeof(line), "bNumInterfaces",
device_name)) {
return -1;
}
if (sscanf(line, "%d", &num_interfaces) != 1) {
return -1;
}
return num_interfaces;
}
static int usb_host_claim_interfaces(USBHostDevice *dev, int configuration)
{
const char *op = NULL;
int dev_descr_len, config_descr_len;
int interface, nb_interfaces;
int ret, i;
for (i = 0; i < USB_MAX_INTERFACES; i++) {
dev->dev.altsetting[i] = 0;
}
if (configuration == 0) { /* address state - ignore */
dev->dev.ninterfaces = 0;
dev->dev.configuration = 0;
return 1;
}
DPRINTF("husb: claiming interfaces. config %d\n", configuration);
i = 0;
dev_descr_len = dev->descr[0];
if (dev_descr_len > dev->descr_len) {
fprintf(stderr, "husb: update iface failed. descr too short\n");
return 0;
}
i += dev_descr_len;
while (i < dev->descr_len) {
DPRINTF("husb: i is %d, descr_len is %d, dl %d, dt %d\n",
i, dev->descr_len,
dev->descr[i], dev->descr[i+1]);
if (dev->descr[i+1] != USB_DT_CONFIG) {
i += dev->descr[i];
continue;
}
config_descr_len = dev->descr[i];
DPRINTF("husb: config #%d need %d\n", dev->descr[i + 5], configuration);
if (configuration == dev->descr[i + 5]) {
configuration = dev->descr[i + 5];
break;
}
i += config_descr_len;
}
if (i >= dev->descr_len) {
fprintf(stderr,
"husb: update iface failed. no matching configuration\n");
return 0;
}
nb_interfaces = dev->descr[i + 4];
if (usb_host_disconnect_ifaces(dev, nb_interfaces) < 0) {
goto fail;
}
/* XXX: only grab if all interfaces are free */
for (interface = 0; interface < nb_interfaces; interface++) {
op = "USBDEVFS_CLAIMINTERFACE";
ret = ioctl(dev->fd, USBDEVFS_CLAIMINTERFACE, &interface);
if (ret < 0) {
goto fail;
}
}
trace_usb_host_claim_interfaces(dev->bus_num, dev->addr,
nb_interfaces, configuration);
dev->dev.ninterfaces = nb_interfaces;
dev->dev.configuration = configuration;
return 1;
fail:
if (errno == ENODEV) {
do_disconnect(dev);
}
perror(op);
return 0;
}
static int usb_host_release_interfaces(USBHostDevice *s)
{
int ret, i;
trace_usb_host_release_interfaces(s->bus_num, s->addr);
for (i = 0; i < s->dev.ninterfaces; i++) {
ret = ioctl(s->fd, USBDEVFS_RELEASEINTERFACE, &i);
if (ret < 0) {
perror("USBDEVFS_RELEASEINTERFACE");
return 0;
}
}
return 1;
}
static void usb_host_handle_reset(USBDevice *dev)
{
USBHostDevice *s = DO_UPCAST(USBHostDevice, dev, dev);
trace_usb_host_reset(s->bus_num, s->addr);
usb_host_do_reset(s);;
usb_host_claim_interfaces(s, 0);
usb_linux_update_endp_table(s);
}
static void usb_host_handle_destroy(USBDevice *dev)
{
USBHostDevice *s = (USBHostDevice *)dev;
usb_host_release_port(s);
usb_host_close(s);
QTAILQ_REMOVE(&hostdevs, s, next);
qemu_remove_exit_notifier(&s->exit);
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
/* iso data is special, we need to keep enough urbs in flight to make sure
that the controller never runs out of them, otherwise the device will
likely suffer a buffer underrun / overrun. */
static AsyncURB *usb_host_alloc_iso(USBHostDevice *s, int pid, uint8_t ep)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
AsyncURB *aurb;
int i, j, len = usb_ep_get_max_packet_size(&s->dev, pid, ep);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
aurb = g_malloc0(s->iso_urb_count * sizeof(*aurb));
for (i = 0; i < s->iso_urb_count; i++) {
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
aurb[i].urb.endpoint = ep;
aurb[i].urb.buffer_length = ISO_FRAME_DESC_PER_URB * len;
aurb[i].urb.buffer = g_malloc(aurb[i].urb.buffer_length);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
aurb[i].urb.type = USBDEVFS_URB_TYPE_ISO;
aurb[i].urb.flags = USBDEVFS_URB_ISO_ASAP;
aurb[i].urb.number_of_packets = ISO_FRAME_DESC_PER_URB;
for (j = 0 ; j < ISO_FRAME_DESC_PER_URB; j++)
aurb[i].urb.iso_frame_desc[j].length = len;
if (pid == USB_TOKEN_IN) {
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
aurb[i].urb.endpoint |= 0x80;
/* Mark as fully consumed (idle) */
aurb[i].iso_frame_idx = ISO_FRAME_DESC_PER_URB;
}
}
set_iso_urb(s, pid, ep, aurb);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
return aurb;
}
static void usb_host_stop_n_free_iso(USBHostDevice *s, int pid, uint8_t ep)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
AsyncURB *aurb;
int i, ret, killed = 0, free = 1;
aurb = get_iso_urb(s, pid, ep);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
if (!aurb) {
return;
}
for (i = 0; i < s->iso_urb_count; i++) {
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
/* in flight? */
if (aurb[i].iso_frame_idx == -1) {
ret = ioctl(s->fd, USBDEVFS_DISCARDURB, &aurb[i]);
if (ret < 0) {
perror("USBDEVFS_DISCARDURB");
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
free = 0;
continue;
}
killed++;
}
}
/* Make sure any urbs we've killed are reaped before we free them */
if (killed) {
async_complete(s);
}
for (i = 0; i < s->iso_urb_count; i++) {
g_free(aurb[i].urb.buffer);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
if (free)
g_free(aurb);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
else
printf("husb: leaking iso urbs because of discard failure\n");
set_iso_urb(s, pid, ep, NULL);
set_iso_urb_idx(s, pid, ep, 0);
clear_iso_started(s, pid, ep);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
static int urb_status_to_usb_ret(int status)
{
switch (status) {
case -EPIPE:
return USB_RET_STALL;
default:
return USB_RET_NAK;
}
}
static int usb_host_handle_iso_data(USBHostDevice *s, USBPacket *p, int in)
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
{
AsyncURB *aurb;
int i, j, ret, max_packet_size, offset, len = 0;
uint8_t *buf;
max_packet_size = p->ep->max_packet_size;
if (max_packet_size == 0)
return USB_RET_NAK;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
aurb = get_iso_urb(s, p->pid, p->ep->nr);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
if (!aurb) {
aurb = usb_host_alloc_iso(s, p->pid, p->ep->nr);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
i = get_iso_urb_idx(s, p->pid, p->ep->nr);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
j = aurb[i].iso_frame_idx;
if (j >= 0 && j < ISO_FRAME_DESC_PER_URB) {
if (in) {
/* Check urb status */
if (aurb[i].urb.status) {
len = urb_status_to_usb_ret(aurb[i].urb.status);
/* Move to the next urb */
aurb[i].iso_frame_idx = ISO_FRAME_DESC_PER_URB - 1;
/* Check frame status */
} else if (aurb[i].urb.iso_frame_desc[j].status) {
len = urb_status_to_usb_ret(
aurb[i].urb.iso_frame_desc[j].status);
/* Check the frame fits */
} else if (aurb[i].urb.iso_frame_desc[j].actual_length
> p->iov.size) {
printf("husb: received iso data is larger then packet\n");
len = USB_RET_NAK;
/* All good copy data over */
} else {
len = aurb[i].urb.iso_frame_desc[j].actual_length;
buf = aurb[i].urb.buffer +
j * aurb[i].urb.iso_frame_desc[0].length;
usb_packet_copy(p, buf, len);
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
} else {
len = p->iov.size;
offset = (j == 0) ? 0 : get_iso_buffer_used(s, p->pid, p->ep->nr);
/* Check the frame fits */
if (len > max_packet_size) {
printf("husb: send iso data is larger then max packet size\n");
return USB_RET_NAK;
}
/* All good copy data over */
usb_packet_copy(p, aurb[i].urb.buffer + offset, len);
aurb[i].urb.iso_frame_desc[j].length = len;
offset += len;
set_iso_buffer_used(s, p->pid, p->ep->nr, offset);
/* Start the stream once we have buffered enough data */
if (!is_iso_started(s, p->pid, p->ep->nr) && i == 1 && j == 8) {
set_iso_started(s, p->pid, p->ep->nr);
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
aurb[i].iso_frame_idx++;
if (aurb[i].iso_frame_idx == ISO_FRAME_DESC_PER_URB) {
i = (i + 1) % s->iso_urb_count;
set_iso_urb_idx(s, p->pid, p->ep->nr, i);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
} else {
if (in) {
set_iso_started(s, p->pid, p->ep->nr);
} else {
DPRINTF("hubs: iso out error no free buffer, dropping packet\n");
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
if (is_iso_started(s, p->pid, p->ep->nr)) {
/* (Re)-submit all fully consumed / filled urbs */
for (i = 0; i < s->iso_urb_count; i++) {
if (aurb[i].iso_frame_idx == ISO_FRAME_DESC_PER_URB) {
ret = ioctl(s->fd, USBDEVFS_SUBMITURB, &aurb[i]);
if (ret < 0) {
perror("USBDEVFS_SUBMITURB");
if (!in || len == 0) {
switch(errno) {
case ETIMEDOUT:
len = USB_RET_NAK;
break;
case EPIPE:
default:
len = USB_RET_STALL;
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
break;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
aurb[i].iso_frame_idx = -1;
change_iso_inflight(s, p->pid, p->ep->nr, 1);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
}
}
return len;
}
static int usb_host_handle_data(USBDevice *dev, USBPacket *p)
{
USBHostDevice *s = DO_UPCAST(USBHostDevice, dev, dev);
struct usbdevfs_urb *urb;
AsyncURB *aurb;
int ret, rem, prem, v;
uint8_t *pbuf;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
uint8_t ep;
trace_usb_host_req_data(s->bus_num, s->addr,
p->pid == USB_TOKEN_IN,
p->ep->nr, p->iov.size);
if (!is_valid(s, p->pid, p->ep->nr)) {
trace_usb_host_req_complete(s->bus_num, s->addr, USB_RET_NAK);
return USB_RET_NAK;
}
if (p->pid == USB_TOKEN_IN) {
ep = p->ep->nr | 0x80;
} else {
ep = p->ep->nr;
}
if (is_halted(s, p->pid, p->ep->nr)) {
unsigned int arg = ep;
ret = ioctl(s->fd, USBDEVFS_CLEAR_HALT, &arg);
if (ret < 0) {
perror("USBDEVFS_CLEAR_HALT");
trace_usb_host_req_complete(s->bus_num, s->addr, USB_RET_NAK);
return USB_RET_NAK;
}
clear_halt(s, p->pid, p->ep->nr);
}
if (is_isoc(s, p->pid, p->ep->nr)) {
return usb_host_handle_iso_data(s, p, p->pid == USB_TOKEN_IN);
}
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
v = 0;
prem = p->iov.iov[v].iov_len;
pbuf = p->iov.iov[v].iov_base;
rem = p->iov.size;
while (rem) {
if (prem == 0) {
v++;
assert(v < p->iov.niov);
prem = p->iov.iov[v].iov_len;
pbuf = p->iov.iov[v].iov_base;
assert(prem <= rem);
}
aurb = async_alloc(s);
aurb->packet = p;
urb = &aurb->urb;
urb->endpoint = ep;
urb->type = usb_host_usbfs_type(s, p);
urb->usercontext = s;
urb->buffer = pbuf;
urb->buffer_length = prem;
if (urb->buffer_length > MAX_USBFS_BUFFER_SIZE) {
urb->buffer_length = MAX_USBFS_BUFFER_SIZE;
}
pbuf += urb->buffer_length;
prem -= urb->buffer_length;
rem -= urb->buffer_length;
if (rem) {
aurb->more = 1;
}
trace_usb_host_urb_submit(s->bus_num, s->addr, aurb,
urb->buffer_length, aurb->more);
ret = ioctl(s->fd, USBDEVFS_SUBMITURB, urb);
DPRINTF("husb: data submit: ep 0x%x, len %u, more %d, packet %p, aurb %p\n",
urb->endpoint, urb->buffer_length, aurb->more, p, aurb);
if (ret < 0) {
perror("USBDEVFS_SUBMITURB");
async_free(aurb);
switch(errno) {
case ETIMEDOUT:
trace_usb_host_req_complete(s->bus_num, s->addr, USB_RET_NAK);
return USB_RET_NAK;
case EPIPE:
default:
trace_usb_host_req_complete(s->bus_num, s->addr, USB_RET_STALL);
return USB_RET_STALL;
}
}
}
return USB_RET_ASYNC;
}
static int ctrl_error(void)
{
if (errno == ETIMEDOUT) {
return USB_RET_NAK;
} else {
return USB_RET_STALL;
}
}
static int usb_host_set_address(USBHostDevice *s, int addr)
{
trace_usb_host_set_address(s->bus_num, s->addr, addr);
s->dev.addr = addr;
return 0;
}
static int usb_host_set_config(USBHostDevice *s, int config)
{
int ret, first = 1;
trace_usb_host_set_config(s->bus_num, s->addr, config);
usb_host_release_interfaces(s);
again:
ret = ioctl(s->fd, USBDEVFS_SETCONFIGURATION, &config);
DPRINTF("husb: ctrl set config %d ret %d errno %d\n", config, ret, errno);
if (ret < 0 && errno == EBUSY && first) {
/* happens if usb device is in use by host drivers */
int count = usb_linux_get_num_interfaces(s);
if (count > 0) {
DPRINTF("husb: busy -> disconnecting %d interfaces\n", count);
usb_host_disconnect_ifaces(s, count);
first = 0;
goto again;
}
}
if (ret < 0) {
return ctrl_error();
}
usb_host_claim_interfaces(s, config);
usb_linux_update_endp_table(s);
return 0;
}
static int usb_host_set_interface(USBHostDevice *s, int iface, int alt)
{
struct usbdevfs_setinterface si;
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
int i, ret;
trace_usb_host_set_interface(s->bus_num, s->addr, iface, alt);
for (i = 1; i <= USB_MAX_ENDPOINTS; i++) {
if (is_isoc(s, USB_TOKEN_IN, i)) {
usb_host_stop_n_free_iso(s, USB_TOKEN_IN, i);
}
if (is_isoc(s, USB_TOKEN_OUT, i)) {
usb_host_stop_n_free_iso(s, USB_TOKEN_OUT, i);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
}
if (iface >= USB_MAX_INTERFACES) {
return USB_RET_STALL;
}
si.interface = iface;
si.altsetting = alt;
ret = ioctl(s->fd, USBDEVFS_SETINTERFACE, &si);
DPRINTF("husb: ctrl set iface %d altset %d ret %d errno %d\n",
iface, alt, ret, errno);
if (ret < 0) {
return ctrl_error();
}
s->dev.altsetting[iface] = alt;
usb_linux_update_endp_table(s);
return 0;
}
static int usb_host_handle_control(USBDevice *dev, USBPacket *p,
int request, int value, int index, int length, uint8_t *data)
{
USBHostDevice *s = DO_UPCAST(USBHostDevice, dev, dev);
struct usbdevfs_urb *urb;
AsyncURB *aurb;
int ret;
/*
* Process certain standard device requests.
* These are infrequent and are processed synchronously.
*/
/* Note request is (bRequestType << 8) | bRequest */
trace_usb_host_req_control(s->bus_num, s->addr, request, value, index);
switch (request) {
case DeviceOutRequest | USB_REQ_SET_ADDRESS:
return usb_host_set_address(s, value);
case DeviceOutRequest | USB_REQ_SET_CONFIGURATION:
return usb_host_set_config(s, value & 0xff);
case InterfaceOutRequest | USB_REQ_SET_INTERFACE:
return usb_host_set_interface(s, index, value);
}
/* The rest are asynchronous */
if (length > sizeof(dev->data_buf)) {
fprintf(stderr, "husb: ctrl buffer too small (%d > %zu)\n",
length, sizeof(dev->data_buf));
return USB_RET_STALL;
}
aurb = async_alloc(s);
aurb->packet = p;
/*
* Setup ctrl transfer.
*
* s->ctrl is laid out such that data buffer immediately follows
* 'req' struct which is exactly what usbdevfs expects.
*/
urb = &aurb->urb;
urb->type = USBDEVFS_URB_TYPE_CONTROL;
urb->endpoint = p->ep->nr;
urb->buffer = &dev->setup_buf;
urb->buffer_length = length + 8;
urb->usercontext = s;
trace_usb_host_urb_submit(s->bus_num, s->addr, aurb,
urb->buffer_length, aurb->more);
ret = ioctl(s->fd, USBDEVFS_SUBMITURB, urb);
DPRINTF("husb: submit ctrl. len %u aurb %p\n", urb->buffer_length, aurb);
if (ret < 0) {
DPRINTF("husb: submit failed. errno %d\n", errno);
async_free(aurb);
switch(errno) {
case ETIMEDOUT:
return USB_RET_NAK;
case EPIPE:
default:
return USB_RET_STALL;
}
}
return USB_RET_ASYNC;
}
static uint8_t usb_linux_get_alt_setting(USBHostDevice *s,
uint8_t configuration, uint8_t interface)
{
char device_name[64], line[1024];
int alt_setting;
sprintf(device_name, "%d-%s:%d.%d", s->bus_num, s->port,
(int)configuration, (int)interface);
if (!usb_host_read_file(line, sizeof(line), "bAlternateSetting",
device_name)) {
/* Assume alt 0 on error */
return 0;
}
if (sscanf(line, "%d", &alt_setting) != 1) {
/* Assume alt 0 on error */
return 0;
}
return alt_setting;
}
/* returns 1 on problem encountered or 0 for success */
static int usb_linux_update_endp_table(USBHostDevice *s)
{
uint8_t *descriptors;
uint8_t devep, type, alt_interface;
uint16_t raw;
int interface, length, i, ep, pid;
struct endp_data *epd;
usb_ep_init(&s->dev);
if (s->dev.configuration == 0) {
/* not configured yet -- leave all endpoints disabled */
return 0;
}
/* get the desired configuration, interface, and endpoint descriptors
* from device description */
descriptors = &s->descr[18];
length = s->descr_len - 18;
i = 0;
while (i < length) {
if (descriptors[i + 1] != USB_DT_CONFIG) {
fprintf(stderr, "invalid descriptor data\n");
return 1;
} else if (descriptors[i + 5] != s->dev.configuration) {
DPRINTF("not requested configuration %d\n", s->dev.configuration);
i += (descriptors[i + 3] << 8) + descriptors[i + 2];
continue;
}
i += descriptors[i];
if (descriptors[i + 1] != USB_DT_INTERFACE ||
(descriptors[i + 1] == USB_DT_INTERFACE &&
descriptors[i + 4] == 0)) {
i += descriptors[i];
continue;
}
interface = descriptors[i + 2];
alt_interface = usb_linux_get_alt_setting(s, s->dev.configuration,
interface);
/* the current interface descriptor is the active interface
* and has endpoints */
if (descriptors[i + 3] != alt_interface) {
i += descriptors[i];
continue;
}
/* advance to the endpoints */
while (i < length && descriptors[i +1] != USB_DT_ENDPOINT) {
i += descriptors[i];
}
if (i >= length)
break;
while (i < length) {
if (descriptors[i + 1] != USB_DT_ENDPOINT) {
break;
}
devep = descriptors[i + 2];
pid = (devep & USB_DIR_IN) ? USB_TOKEN_IN : USB_TOKEN_OUT;
ep = devep & 0xf;
if (ep == 0) {
fprintf(stderr, "usb-linux: invalid ep descriptor, ep == 0\n");
return 1;
}
type = descriptors[i + 3] & 0x3;
raw = descriptors[i + 4] + (descriptors[i + 5] << 8);
usb_ep_set_max_packet_size(&s->dev, pid, ep, raw);
assert(usb_ep_get_type(&s->dev, pid, ep) ==
USB_ENDPOINT_XFER_INVALID);
usb_ep_set_type(&s->dev, pid, ep, type);
usb_ep_set_ifnum(&s->dev, pid, ep, interface);
epd = get_endp(s, pid, ep);
epd->halted = 0;
i += descriptors[i];
}
}
#ifdef DEBUG
usb_ep_dump(&s->dev);
#endif
return 0;
}
/*
* Check if we can safely redirect a usb2 device to a usb1 virtual controller,
* this function assumes this is safe, if:
* 1) There are no isoc endpoints
* 2) There are no interrupt endpoints with a max_packet_size > 64
* Note bulk endpoints with a max_packet_size > 64 in theory also are not
* usb1 compatible, but in practice this seems to work fine.
*/
static int usb_linux_full_speed_compat(USBHostDevice *dev)
{
int i, packet_size;
/*
* usb_linux_update_endp_table only registers info about ep in the current
* interface altsettings, so we need to parse the descriptors again.
*/
for (i = 0; (i + 5) < dev->descr_len; i += dev->descr[i]) {
if (dev->descr[i + 1] == USB_DT_ENDPOINT) {
switch (dev->descr[i + 3] & 0x3) {
case 0x00: /* CONTROL */
break;
case 0x01: /* ISO */
return 0;
case 0x02: /* BULK */
break;
case 0x03: /* INTERRUPT */
packet_size = dev->descr[i + 4] + (dev->descr[i + 5] << 8);
if (packet_size > 64)
return 0;
break;
}
}
}
return 1;
}
static int usb_host_open(USBHostDevice *dev, int bus_num,
int addr, const char *port,
const char *prod_name, int speed)
{
int fd = -1, ret;
trace_usb_host_open_started(bus_num, addr);
if (dev->fd != -1) {
goto fail;
}
fd = usb_host_open_device(bus_num, addr);
if (fd < 0) {
goto fail;
}
DPRINTF("husb: opened %s\n", buf);
dev->bus_num = bus_num;
dev->addr = addr;
strcpy(dev->port, port);
dev->fd = fd;
/* read the device description */
dev->descr_len = read(fd, dev->descr, sizeof(dev->descr));
if (dev->descr_len <= 0) {
perror("husb: reading device data failed");
goto fail;
}
#ifdef DEBUG
{
int x;
printf("=== begin dumping device descriptor data ===\n");
for (x = 0; x < dev->descr_len; x++) {
printf("%02x ", dev->descr[x]);
}
printf("\n=== end dumping device descriptor data ===\n");
}
#endif
/* start unconfigured -- we'll wait for the guest to set a configuration */
if (!usb_host_claim_interfaces(dev, 0)) {
goto fail;
}
ret = usb_linux_update_endp_table(dev);
if (ret) {
goto fail;
}
if (speed == -1) {
struct usbdevfs_connectinfo ci;
ret = ioctl(fd, USBDEVFS_CONNECTINFO, &ci);
if (ret < 0) {
perror("usb_host_device_open: USBDEVFS_CONNECTINFO");
goto fail;
}
if (ci.slow) {
speed = USB_SPEED_LOW;
} else {
speed = USB_SPEED_HIGH;
}
}
dev->dev.speed = speed;
dev->dev.speedmask = (1 << speed);
if (dev->dev.speed == USB_SPEED_HIGH && usb_linux_full_speed_compat(dev)) {
dev->dev.speedmask |= USB_SPEED_MASK_FULL;
}
trace_usb_host_open_success(bus_num, addr);
if (!prod_name || prod_name[0] == '\0') {
snprintf(dev->dev.product_desc, sizeof(dev->dev.product_desc),
"host:%d.%d", bus_num, addr);
} else {
pstrcpy(dev->dev.product_desc, sizeof(dev->dev.product_desc),
prod_name);
}
ret = usb_device_attach(&dev->dev);
if (ret) {
goto fail;
}
/* USB devio uses 'write' flag to check for async completions */
qemu_set_fd_handler(dev->fd, NULL, async_complete, dev);
return 0;
fail:
trace_usb_host_open_failure(bus_num, addr);
if (dev->fd != -1) {
close(dev->fd);
dev->fd = -1;
}
return -1;
}
static int usb_host_close(USBHostDevice *dev)
{
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
int i;
if (dev->fd == -1) {
return -1;
}
trace_usb_host_close(dev->bus_num, dev->addr);
qemu_set_fd_handler(dev->fd, NULL, NULL, NULL);
dev->closing = 1;
for (i = 1; i <= USB_MAX_ENDPOINTS; i++) {
if (is_isoc(dev, USB_TOKEN_IN, i)) {
usb_host_stop_n_free_iso(dev, USB_TOKEN_IN, i);
}
if (is_isoc(dev, USB_TOKEN_OUT, i)) {
usb_host_stop_n_free_iso(dev, USB_TOKEN_OUT, i);
usb-linux: Add support for buffering iso usb packets Currently we are submitting iso packets to the host one at a time, as we receive them from the emulated host controller. This has 2 problems: 1) If we were fast enough to submit every packet in time for the next host host controller usb frame, we would be generating 1000 hardware interrupts per second on the host 2) We are not fast enough to submit every packet in time for the next host host controller usb frame, causing us to not submit iso urbs in some usb frames which causes devices with an endpoint with an interval of 1 ms (so every frame) to loose data. This causes for example ubs-1.1 webcams to not work properly (usb-2.0 is not supported at all atm). This patch fixes both problems by changing the iso packet pass through handling to buffer packets. This version only does so for iso input packets (webcams, audio in) I'm working on a second patch extending this to iso output packets (audio out). This patch makes use of the linux batching of iso packets in one urb. When an iso in packet gets received from the emulated host controller, it immediately submits 3 urbs with 32 iso in packets each. This causes the host to only get an hw interrupt every 32 packets dropping the interrupt rate to 32 interrupts per second and gives it a queue of urbs to work from once the first 32 iso in packets have been received to make sure no packets are dropped. Besides submitting a whole bunch or urbs as soon as the first urb is received, effectively creating a buffer inside the kernel, this patch also gets rid of the asynchroneous completion for iso in urbs. Instead they are only marked as complete in the fd write callback (which usbfs uses to signal complete urbs). These complete packets then get consumed by returning them synchroneously to the emulated host controller when it submits an iso in packet for the ep in question. When no complete packets are ready (which happens when the stream is starting) a 0 length packet gets returned to the emulated host controller. With this patch I've several usb-1.1 webcams working well with usb pass through, where as without this patch none of them work. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
2010-11-26 13:41:08 +03:00
}
}
async_complete(dev);
dev->closing = 0;
if (dev->dev.attached) {
usb_device_detach(&dev->dev);
}
usb_host_do_reset(dev);
close(dev->fd);
dev->fd = -1;
return 0;
}
static void usb_host_exit_notifier(struct Notifier *n, void *data)
{
USBHostDevice *s = container_of(n, USBHostDevice, exit);
usb_host_release_port(s);
if (s->fd != -1) {
usb_host_do_reset(s);;
}
}
static int usb_host_initfn(USBDevice *dev)
{
USBHostDevice *s = DO_UPCAST(USBHostDevice, dev, dev);
dev->auto_attach = 0;
s->fd = -1;
s->hub_fd = -1;
QTAILQ_INSERT_TAIL(&hostdevs, s, next);
s->exit.notify = usb_host_exit_notifier;
qemu_add_exit_notifier(&s->exit);
usb_host_auto_check(NULL);
if (s->match.bus_num != 0 && s->match.port != NULL) {
usb_host_claim_port(s);
}
return 0;
}
static const VMStateDescription vmstate_usb_host = {
.name = "usb-host",
.unmigratable = 1,
};
static Property usb_host_dev_properties[] = {
DEFINE_PROP_UINT32("hostbus", USBHostDevice, match.bus_num, 0),
DEFINE_PROP_UINT32("hostaddr", USBHostDevice, match.addr, 0),
DEFINE_PROP_STRING("hostport", USBHostDevice, match.port),
DEFINE_PROP_HEX32("vendorid", USBHostDevice, match.vendor_id, 0),
DEFINE_PROP_HEX32("productid", USBHostDevice, match.product_id, 0),
DEFINE_PROP_UINT32("isobufs", USBHostDevice, iso_urb_count, 4),
DEFINE_PROP_END_OF_LIST(),
};
static void usb_host_class_initfn(ObjectClass *klass, void *data)
{
DeviceClass *dc = DEVICE_CLASS(klass);
USBDeviceClass *uc = USB_DEVICE_CLASS(klass);
uc->init = usb_host_initfn;
uc->product_desc = "USB Host Device";
uc->cancel_packet = usb_host_async_cancel;
uc->handle_data = usb_host_handle_data;
uc->handle_control = usb_host_handle_control;
uc->handle_reset = usb_host_handle_reset;
uc->handle_destroy = usb_host_handle_destroy;
dc->vmsd = &vmstate_usb_host;
dc->props = usb_host_dev_properties;
}
static TypeInfo usb_host_dev_info = {
.name = "usb-host",
.parent = TYPE_USB_DEVICE,
.instance_size = sizeof(USBHostDevice),
.class_init = usb_host_class_initfn,
};
static void usb_host_register_types(void)
{
type_register_static(&usb_host_dev_info);
usb_legacy_register("usb-host", "host", usb_host_device_open);
}
type_init(usb_host_register_types)
USBDevice *usb_host_device_open(const char *devname)
{
struct USBAutoFilter filter;
USBDevice *dev;
char *p;
dev = usb_create(NULL /* FIXME */, "usb-host");
if (strstr(devname, "auto:")) {
if (parse_filter(devname, &filter) < 0) {
goto fail;
}
} else {
if ((p = strchr(devname, '.'))) {
filter.bus_num = strtoul(devname, NULL, 0);
filter.addr = strtoul(p + 1, NULL, 0);
filter.vendor_id = 0;
filter.product_id = 0;
} else if ((p = strchr(devname, ':'))) {
filter.bus_num = 0;
filter.addr = 0;
filter.vendor_id = strtoul(devname, NULL, 16);
filter.product_id = strtoul(p + 1, NULL, 16);
} else {
goto fail;
}
}
qdev_prop_set_uint32(&dev->qdev, "hostbus", filter.bus_num);
qdev_prop_set_uint32(&dev->qdev, "hostaddr", filter.addr);
qdev_prop_set_uint32(&dev->qdev, "vendorid", filter.vendor_id);
qdev_prop_set_uint32(&dev->qdev, "productid", filter.product_id);
qdev_init_nofail(&dev->qdev);
return dev;
fail:
qdev_free(&dev->qdev);
return NULL;
}
int usb_host_device_close(const char *devname)
{
#if 0
char product_name[PRODUCT_NAME_SZ];
int bus_num, addr;
USBHostDevice *s;
if (strstr(devname, "auto:")) {
return usb_host_auto_del(devname);
}
if (usb_host_find_device(&bus_num, &addr, product_name,
sizeof(product_name), devname) < 0) {
return -1;
}
s = hostdev_find(bus_num, addr);
if (s) {
usb_device_delete_addr(s->bus_num, s->dev.addr);
return 0;
}
#endif
return -1;
}
/*
* Read sys file-system device file
*
* @line address of buffer to put file contents in
* @line_size size of line
* @device_file path to device file (printf format string)
* @device_name device being opened (inserted into device_file)
*
* @return 0 failed, 1 succeeded ('line' contains data)
*/
static int usb_host_read_file(char *line, size_t line_size,
const char *device_file, const char *device_name)
{
FILE *f;
int ret = 0;
char filename[PATH_MAX];
snprintf(filename, PATH_MAX, "/sys/bus/usb/devices/%s/%s", device_name,
device_file);
f = fopen(filename, "r");
if (f) {
ret = fgets(line, line_size, f) != NULL;
fclose(f);
}
return ret;
}
/*
* Use /sys/bus/usb/devices/ directory to determine host's USB
* devices.
*
* This code is based on Robert Schiele's original patches posted to
* the Novell bug-tracker https://bugzilla.novell.com/show_bug.cgi?id=241950
*/
static int usb_host_scan(void *opaque, USBScanFunc *func)
{
DIR *dir = NULL;
char line[1024];
int bus_num, addr, speed, class_id, product_id, vendor_id;
int ret = 0;
char port[MAX_PORTLEN];
char product_name[512];
struct dirent *de;
dir = opendir("/sys/bus/usb/devices");
if (!dir) {
perror("husb: opendir /sys/bus/usb/devices");
fprintf(stderr, "husb: please make sure sysfs is mounted at /sys\n");
goto the_end;
}
while ((de = readdir(dir))) {
if (de->d_name[0] != '.' && !strchr(de->d_name, ':')) {
if (sscanf(de->d_name, "%d-%7[0-9.]", &bus_num, port) < 2) {
continue;
}
if (!usb_host_read_file(line, sizeof(line), "devnum", de->d_name)) {
goto the_end;
}
if (sscanf(line, "%d", &addr) != 1) {
goto the_end;
}
if (!usb_host_read_file(line, sizeof(line), "bDeviceClass",
de->d_name)) {
goto the_end;
}
if (sscanf(line, "%x", &class_id) != 1) {
goto the_end;
}
if (!usb_host_read_file(line, sizeof(line), "idVendor",
de->d_name)) {
goto the_end;
}
if (sscanf(line, "%x", &vendor_id) != 1) {
goto the_end;
}
if (!usb_host_read_file(line, sizeof(line), "idProduct",
de->d_name)) {
goto the_end;
}
if (sscanf(line, "%x", &product_id) != 1) {
goto the_end;
}
if (!usb_host_read_file(line, sizeof(line), "product",
de->d_name)) {
*product_name = 0;
} else {
if (strlen(line) > 0) {
line[strlen(line) - 1] = '\0';
}
pstrcpy(product_name, sizeof(product_name), line);
}
if (!usb_host_read_file(line, sizeof(line), "speed", de->d_name)) {
goto the_end;
}
if (!strcmp(line, "5000\n")) {
speed = USB_SPEED_SUPER;
} else if (!strcmp(line, "480\n")) {
speed = USB_SPEED_HIGH;
} else if (!strcmp(line, "1.5\n")) {
speed = USB_SPEED_LOW;
} else {
speed = USB_SPEED_FULL;
}
ret = func(opaque, bus_num, addr, port, class_id, vendor_id,
product_id, product_name, speed);
if (ret) {
goto the_end;
}
}
}
the_end:
if (dir) {
closedir(dir);
}
return ret;
}
static QEMUTimer *usb_auto_timer;
static int usb_host_auto_scan(void *opaque, int bus_num,
int addr, const char *port,
int class_id, int vendor_id, int product_id,
const char *product_name, int speed)
{
struct USBAutoFilter *f;
struct USBHostDevice *s;
/* Ignore hubs */
if (class_id == 9)
return 0;
QTAILQ_FOREACH(s, &hostdevs, next) {
f = &s->match;
if (f->bus_num > 0 && f->bus_num != bus_num) {
continue;
}
if (f->addr > 0 && f->addr != addr) {
continue;
}
if (f->port != NULL && (port == NULL || strcmp(f->port, port) != 0)) {
continue;
}
if (f->vendor_id > 0 && f->vendor_id != vendor_id) {
continue;
}
if (f->product_id > 0 && f->product_id != product_id) {
continue;
}
/* We got a match */
s->seen++;
if (s->errcount >= 3) {
return 0;
}
/* Already attached ? */
if (s->fd != -1) {
return 0;
}
DPRINTF("husb: auto open: bus_num %d addr %d\n", bus_num, addr);
if (usb_host_open(s, bus_num, addr, port, product_name, speed) < 0) {
s->errcount++;
}
break;
}
return 0;
}
static void usb_host_auto_check(void *unused)
{
struct USBHostDevice *s;
int unconnected = 0;
usb_host_scan(NULL, usb_host_auto_scan);
QTAILQ_FOREACH(s, &hostdevs, next) {
if (s->fd == -1) {
unconnected++;
}
if (s->seen == 0) {
s->errcount = 0;
}
s->seen = 0;
}
if (unconnected == 0) {
/* nothing to watch */
if (usb_auto_timer) {
qemu_del_timer(usb_auto_timer);
trace_usb_host_auto_scan_disabled();
}
return;
}
if (!usb_auto_timer) {
usb_auto_timer = qemu_new_timer_ms(rt_clock, usb_host_auto_check, NULL);
if (!usb_auto_timer) {
return;
}
trace_usb_host_auto_scan_enabled();
}
qemu_mod_timer(usb_auto_timer, qemu_get_clock_ms(rt_clock) + 2000);
}
/*
* Autoconnect filter
* Format:
* auto:bus:dev[:vid:pid]
* auto:bus.dev[:vid:pid]
*
* bus - bus number (dec, * means any)
* dev - device number (dec, * means any)
* vid - vendor id (hex, * means any)
* pid - product id (hex, * means any)
*
* See 'lsusb' output.
*/
static int parse_filter(const char *spec, struct USBAutoFilter *f)
{
enum { BUS, DEV, VID, PID, DONE };
const char *p = spec;
int i;
f->bus_num = 0;
f->addr = 0;
f->vendor_id = 0;
f->product_id = 0;
for (i = BUS; i < DONE; i++) {
p = strpbrk(p, ":.");
if (!p) {
break;
}
p++;
if (*p == '*') {
continue;
}
switch(i) {
case BUS: f->bus_num = strtol(p, NULL, 10); break;
case DEV: f->addr = strtol(p, NULL, 10); break;
case VID: f->vendor_id = strtol(p, NULL, 16); break;
case PID: f->product_id = strtol(p, NULL, 16); break;
}
}
if (i < DEV) {
fprintf(stderr, "husb: invalid auto filter spec %s\n", spec);
return -1;
}
return 0;
}
/**********************/
/* USB host device info */
struct usb_class_info {
int class;
const char *class_name;
};
static const struct usb_class_info usb_class_info[] = {
{ USB_CLASS_AUDIO, "Audio"},
{ USB_CLASS_COMM, "Communication"},
{ USB_CLASS_HID, "HID"},
{ USB_CLASS_HUB, "Hub" },
{ USB_CLASS_PHYSICAL, "Physical" },
{ USB_CLASS_PRINTER, "Printer" },
{ USB_CLASS_MASS_STORAGE, "Storage" },
{ USB_CLASS_CDC_DATA, "Data" },
{ USB_CLASS_APP_SPEC, "Application Specific" },
{ USB_CLASS_VENDOR_SPEC, "Vendor Specific" },
{ USB_CLASS_STILL_IMAGE, "Still Image" },
{ USB_CLASS_CSCID, "Smart Card" },
{ USB_CLASS_CONTENT_SEC, "Content Security" },
{ -1, NULL }
};
static const char *usb_class_str(uint8_t class)
{
const struct usb_class_info *p;
for(p = usb_class_info; p->class != -1; p++) {
if (p->class == class) {
break;
}
}
return p->class_name;
}
static void usb_info_device(Monitor *mon, int bus_num,
int addr, const char *port,
int class_id, int vendor_id, int product_id,
const char *product_name,
int speed)
{
const char *class_str, *speed_str;
switch(speed) {
case USB_SPEED_LOW:
speed_str = "1.5";
break;
case USB_SPEED_FULL:
speed_str = "12";
break;
case USB_SPEED_HIGH:
speed_str = "480";
break;
case USB_SPEED_SUPER:
speed_str = "5000";
break;
default:
speed_str = "?";
break;
}
monitor_printf(mon, " Bus %d, Addr %d, Port %s, Speed %s Mb/s\n",
bus_num, addr, port, speed_str);
class_str = usb_class_str(class_id);
if (class_str) {
monitor_printf(mon, " %s:", class_str);
} else {
monitor_printf(mon, " Class %02x:", class_id);
}
monitor_printf(mon, " USB device %04x:%04x", vendor_id, product_id);
if (product_name[0] != '\0') {
monitor_printf(mon, ", %s", product_name);
}
monitor_printf(mon, "\n");
}
static int usb_host_info_device(void *opaque, int bus_num, int addr,
const char *path, int class_id,
int vendor_id, int product_id,
const char *product_name,
int speed)
{
Monitor *mon = opaque;
usb_info_device(mon, bus_num, addr, path, class_id, vendor_id, product_id,
product_name, speed);
return 0;
}
static void dec2str(int val, char *str, size_t size)
{
if (val == 0) {
snprintf(str, size, "*");
} else {
snprintf(str, size, "%d", val);
}
}
static void hex2str(int val, char *str, size_t size)
{
if (val == 0) {
snprintf(str, size, "*");
} else {
snprintf(str, size, "%04x", val);
}
}
void usb_host_info(Monitor *mon)
{
struct USBAutoFilter *f;
struct USBHostDevice *s;
usb_host_scan(mon, usb_host_info_device);
if (QTAILQ_EMPTY(&hostdevs)) {
return;
}
monitor_printf(mon, " Auto filters:\n");
QTAILQ_FOREACH(s, &hostdevs, next) {
char bus[10], addr[10], vid[10], pid[10];
f = &s->match;
dec2str(f->bus_num, bus, sizeof(bus));
dec2str(f->addr, addr, sizeof(addr));
hex2str(f->vendor_id, vid, sizeof(vid));
hex2str(f->product_id, pid, sizeof(pid));
monitor_printf(mon, " Bus %s, Addr %s, Port %s, ID %s:%s\n",
bus, addr, f->port ? f->port : "*", vid, pid);
}
}