We already have USB_RET_NAK, but that means that a device does not want
to send/receive right now. But with host / network redirection we can
actually have a transaction fail due to some io error, rather then ie
the device just not having any data atm.
This patch adds a new error code named USB_RET_IOERROR for this, and uses
it were appropriate.
Notes:
-Currently all usb-controllers handle this the same as NODEV, but that
may change in the future, OHCI could indicate a CRC error instead for example.
-This patch does not touch hw/usb-musb.c, that is because the code in there
handles STALL and NAK specially and has a if status < 0 generic catch all
for all other errors
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
All error statuses except for NAK are handled in a switch case, move the
handling of NAK into the same switch case.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The nakcnt code in ehci_execute_complete() marked transactions as finished
when a packet completed with a result of USB_RET_NAK, but USB_RET_NAK
means that the device cannot receive / send data at that time and that
the transaction should be retried later, which is also what the usb-uhci
and usb-ohci code does.
Note that there already was some special code in place to handle this
for interrupt endpoints in the form of doing a return from
ehci_execute_complete() when reload == 0, but that for bulk transactions
this was not handled correctly (where as for example the usb-ccid device does
return USB_RET_NAK for bulk packets).
Besides that the code in ehci_execute_complete() decrement nakcnt by 1
on a packet result of USB_RET_NAK, but
-since the transaction got marked as finished,
nakcnt would never be decremented again
-there is no code checking for nakcnt becoming 0
-there is no use in re-trying the transaction within the same usb frame /
usb-ehci frame-timer call, since the status of emulated devices won't change
as long as the usb-ehci frame-timer is running
So we should simply set the nakcnt to 0 when we get a USB_RET_NAK, thus
claiming that we've tried reload times (or as many times as possible if
reload is 0).
Besides the code in ehci_execute_complete() handling USB_RET_NAK there
was also code handling it in ehci_state_executing(), which calls
ehci_execute_complete(), and then does its own handling on top of the handling
in ehci_execute_complete(), this code would decrement nakcnt *again* (if not
already 0), or restore the reload value (which was never changed) on success.
Since the double decrement was wrong to begin with, and is no longer needed
now that we set nakcnt directly to 0 on USB_RET_NAK, and the restore of reload
is not needed either, this patch simply removes all nakcnt handling from
ehci_state_executing().
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This patch removes 2 bits of dead nakcnt code:
1) usb_ehci_execute calls ehci_qh_do_overlay which does:
nakcnt = reload;
and then has a block of code which is conditional on:
if (reload && !nakcnt) {
which ofcourse is never true now as nakcnt == reload.
2) ehci_state_fetchqh does:
nakcnt = reload;
but before nakcnt is ever used ehci_state_fetchqh is always followed
by a ehci_qh_do_overlay call which also does:
nakcnt = reload;
So doing this from ehci_state_fetchqh is redundant.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
cerr should only be decremented on errors which cause XactErr to be set, and
when that happens the failing transaction should be retried until cerr reaches
0 and only then should USBSTS_ERRINT be set (and inactive cleared and
USBSTS_INT set if requested).
Since we don't have any hardware level errors (and in case of redirection
the real hardware has already retried), re-trying makes no sense, so
immediately set cerr to 0 on errors which set XactErr.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
As clearly stated in the 2.3.2 of the EHCI spec, any time USBERRINT get
sets then if the td has its IOC bit set USBINT should be set as well.
This means that for any status except for USB_RET_NAK we should set
USBINT if the IOC bit is set.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The purpose of the IAAD bit / the doorbell is to make the ehci controller
forget about cached qhs, this is mainly used when cancelling transactions,
the qh is unlinked from the async schedule and then the doorbell gets rung,
once the doorbell is acked by the controller the hcd knows that the qh is
no longer in use and that it can do something else with the memory, such
as re-use it for a new qh! But we keep our struct representing this qh around
for circa 250 ms. This allows for a (mightily large) race window where the
following could happen:
-hcd submits a qh at address 0xdeadbeef
-our ehci code sees the qh, sends a request to a usb-device, gets a result
of USB_RET_ASYNC, sets the async_state of the qh to EHCI_ASYNC_INFLIGHT
-hcd unlinks the qh at address 0xdeadbeef
-hcd rings the doorbell, wait for us to ack it
-hcd re-uses the qh at address 0xdeadbeef
-our ehci code sees the qh, looks in the async_queue, sees there already is
a qh at address 0xdeadbeef there with async_state of EHCI_ASYNC_INFLIGHT,
does nothing
-the *original* (which the hcd thinks it has cancelled) transaction finishes
-our ehci code sees the qh on yet another pass through the async list,
looks in the async_queue, sees there already is a qh at address 0xdeadbeef
there with async_state of EHCI_ASYNC_COMPLETED, and finished the transaction
with the results of the *original* transaction.
Not good (tm), this patch fixes this race by removing all qhs which have not
been seen during the last cycle through the async list immidiately when the
doorbell is rung.
Note this patch does not fix any actually observed problem, but upon
reading of the EHCI spec it became apparent to me that the above race could
happen and the usb-ehci behavior from before this patch is not good.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Before this patch USB 2 devices with interrupt endpoints were not working
properly. The problem is that to avoid loops we stop processing as soon
as we encounter a queue-head (qh) we've already seen since qhs can be linked
in a circular fashion, this is tracked by the seen flag in our qh struct.
The resetting of the seen flag is done from ehci_queues_rip_unused which
before this patch was only called when executing the statemachine for the
async schedule.
But packets for interrupt endpoints are part of the periodic schedule! So what
would happen is that when there were no ctrl or bulk packets for a USB 2
device with an interrupt endpoint, the async schedule would become non
active, then ehci_queues_rip_unused would no longer get called and when
processing the qhs for the interrupt endpoints from the periodic schedule
their seen bit would still be 1 and they would be skipped.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
qhs can be part of both the async and the periodic schedule, as is shown
in later patches in this series it is useful to keep track of the qhs on
a per schedule basis.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Before this patch the T-bit was not checked in 2 places, while it should be.
Once we properly check the T-bit everywhere we no longer need the weird
entry < 0x1000 and entry > 0x1000 checks, so this patch removes them.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Before this patch the ehci code was not checking for any other errors other
then USB_RET_NAK. This causes 2 problems:
1) Other errors are not reported to the guest.
2) When transactions with the ITD_XACT_IOC bit set completing with another
error would not result in USBSTS_INT getting set.
I hit this problem when unplugging devices while iso data was streaming from
the device to the guest. When this happens it takes a while for the guest to
process the unplugging and remove ISO transactions from the ehci schedule, in
the mean time these transactions would complete with a result of USB_RET_NODEV,
which was not handled. This lead to the Linux guest's usb subsystem "hanging",
that is it would no longer see new usb devices getting plugged in and running
for example lsusb would lead to a stuck (D state) lsusb process. This patch
fixes this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This patch adds a sanity check to itd processing to make sure the
endpoint addressed by the guest is actually an iso endpoint. Also
verify that usb drivers don't return USB_RET_ASYNC which is illegal for
iso xfers.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Two reset fixes:
* pick up s->usbcmd value after ehci_reset call to make sure it
keeps the reset value and doesn't get rubbish filled in when
val is written back to the mmio register array later on.
* make sure the frame timer is zapped on reset.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
* kraxel/usb.38: (28 commits)
xhci: handle USB_RET_NAK
xhci: remote wakeup support
xhci: kill port arg from xhci_setup_packet
xhci: stop on errors
xhci: add trb type name lookup support.
xhci: signal low- and fullspeed support
usb: add USBBusOps->wakeup_endpoint
usb: pass USBEndpoint to usb_wakeup
usb: maintain async packet list per endpoint
usb: Set USBEndpoint in usb_packet_setup().
usb: add USBEndpoint->{nr,pid}
usb: USBPacket: add status, rename owner -> ep
usb: fold usb_generic_handle_packet into usb_handle_packet
usb: kill handle_packet callback
usb-xhci: switch to usb_find_device()
usb-musb: switch to usb_find_device()
usb-ohci: switch to usb_find_device()
usb-ehci: switch to usb_find_device()
usb-uhci: switch to usb_find_device()
usb: handle dev == NULL in usb_handle_packet()
...
Replace device_init() with generalized type_init().
While at it, unify naming convention: type_init([$prefix_]register_types)
Also, type_init() is a function, so add preceding blank line where
necessary and don't put a semicolon after the closing brace.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Anthony Liguori <anthony@codemonkey.ws>
Cc: malc <av1474@comtv.ru>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
With the separation of the device lookup (via usb_find_device) and
packet processing we can lookup device and endpoint before setting up
the usb packet. So we can initialize USBPacket->ep early and keep it
valid for the whole lifecycle of the USBPacket. Also the devaddr and
devep fields are not needed any more.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Add enum to track the status of USBPackets, use that instead of the
owner pointer to figure whenever a usb packet is currently in flight
or not. Add some more packet status sanity checks. Also rename the
USBEndpoint pointer from "owner" to "ep".
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The USB subsystem pipes internal reset notifications through
usb_handle_packet() with a special magic PID. This indirection
is a pretty pointless excercise as it ends up being handled by
usb_generic_handle_packet anyway.
Replace the USB_MSG_RESET with a usb_device_reset() function
which can be called directly. Also rename the existing usb_reset()
function to usb_port_reset() to avoid confusion.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
According to the EHCI spec port ownership should revert to the EHCI controller
on device disconnect. This fixes the problem of a port getting stuck on USB 1
when using redirection and plugging in a USB 2 device after a USB 1 device
has been redirected.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This was done in a mostly automated fashion. I did it in three steps and then
rebased it into a single step which avoids repeatedly touching every file in
the tree.
The first step was a sed-based addition of the parent type to the subclass
registration functions.
The second step was another sed-based removal of subclass registration functions
while also adding virtual functions from the base class into a class_init
function as appropriate.
Finally, a python script was used to convert the DeviceInfo structures and
qdev_register_subclass functions to TypeInfo structures, class_init functions,
and type_register_static calls.
We are almost fully converted to QOM after this commit.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
These are various small stylistic changes which help make things more
consistent such that the automated conversion script can be simpler.
It's not necessary to agree or disagree with these style changes because all
of this code is going to be rewritten by the patch monkey script anyway.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Add USBEndpoint for the control endpoint to USBDevices. Link async
packets to the USBEndpoint instead of the USBDevice.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This patch removes some unnecessary casts in the usb-ehci device,
introduced by commit 68d553587c
'usb-ehci: Use PCI DMA stub functions'.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The mmio register name list only had the names for four port status
registers. We emulate a EHCI adapter with six ports though, the last
two ones are listed as "unknown" in traces. Fix it.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Coverity thinks q could be NULL there and warns.
I believe it can't be NULL there.
Add assert to prove it.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This updates the usb-ehci device emulation to use the explicit PCI DMA
wrapper to initialize its scatter/gathjer structure. This means this
driver should not need further changes when the sglist interface is
extended to support IOMMUs.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
commit 891fb2cd45 removed the implicit
detach before (re-)attaching in usb_attach(). Some usb host controllers
used that behavior though to do a port reset by a detach+attach
sequence.
This patch establishes old behavior by adding a new usb_reset() function
for port resets and putting it into use, thereby also unifying port
reset behavior of all host controllers. The patch also adds asserts to
usb_attach() and usb_detach() to make sure the calls are symmetrical.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Fix up some erroneous comments in code:
interrupt pins are named A-D, the
interrupt pin register is always readonly
and isn't zeroed out on reset.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch makes qemu assign a port when creating the device, not when
attaching it. For most usb devices this isn't a noticable difference
because they are in attached state all the time.
The change affects usb-host devices which live in detached state while
the real device is unplugged from the host. They have a fixed port
assigned all the time now instead of getting grabbing one on attach and
releasing it at detach, i.e. they stop floating around at the usb bus.
The change also allows to simplify usb-hub. It doesn't need the
handle_attach() callback any more to configure the downstream ports.
This can be done at device initialitation time now. The changed
initialization order (first grab upstream port, then register downstream
ports) also fixes some icky corner cases. For example it is not possible
any more to plug the hub into one of its own downstream ports.
The usb host adapters must care too. USBPort->dev being non-NULL
doesn't imply any more the device is in attached state. The host
adapters must additionally check the USBPort->dev->attached flag.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This patch adds code to do minimal siTD handling, which is basically
just following the next pointer. This is good enougth to handle the
inactive siTDs used by FreeBSD. Active siTDs are skipped too as we
don't have split transfer support in qemu, additionally a warning is
printed.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Map guest memory and pass on a direct pointer instead of copying
the bits to a indirect buffer. EHCI transfer descriptors can
reference multiple (physical guest) pages so we'll actually start
seeing usb packets wich carry iovec with more than one element.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Zap data pointer from USBPacket, add a QEMUIOVector instead.
Add a bunch of helper functions to manage USBPacket data.
Switch over users to the new interface.
Note that USBPacket->len was used for two purposes: First to
pass in the buffer size and second to return the number of
transfered bytes or the status code on async transfers. There
is a new result variable for the latter. A new status code
was added to catch uninitialized result.
Nobody creates iovecs with more than one element (yet).
Some users are (temporarely) limited to iovecs with a single
element to keep the patch size as small as possible.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The PED bit should only be set for highspeed devices and the PEDC bit
should not be set on "normal" PED bit changes, only on io errors.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Doing a usb_attach when dev is NULL will just result in the
port detach op getting called even though nothing was connected in
the first place.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Note this fixes 2 things in one go, first of all the device_destroy bus
op should be a device_detach bus op, as pending async packets from the
device should be cancelled on detach not on destroy.
Secondly having this as a bus op won't work with companion controllers, since
then there will be 1 bus driven by the ehci controller and thus 1 set of bus
ops, but the device being detached may be downstream of a handed over port.
Making the detach of a downstream device a port op allows the ehci controller
to forward this to the companion controller port for handed over ports.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This makes them consistent with the attach and detach ops, and in general
it makes sense to make portops take a port as argument. This also makes
adding support for a companion controller easier / cleaner.
[ kraxel: fix usb-musb.c build ]
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
