on current-users, with cross-check and some improvement from linux-2.4.19
and FreeBSD-current.
Also don't set the APO_UDMA_CLK66 bit for Ultra/100 capable chipset, and
support Ultra/133 for the VT8233A.
sync that descriptor with PREREAD to make sure that it is evicted
from the data cache. From Allen Briggs.
* With the above bug fixed, stop using BUS_DMA_COHERENT, resulting in
a fairly decent performance improvement on systems where BUS_DMA_COHERENT
causes descriptors to be accessed uncached (most painful in wm_start()).
-Free AGP memory on close, to avoid a memory leak in case
the X server doesn't free it explicitely.
(It appears that the X server never calls AGPIOC_DEALLOCATE.)
Fixes PR kern/17869 by Emmanuel Dreyfus.
i82546 chip has 2 Gig-E interfaces in a single chip.
The main difference is these chips have a special handshaking protocol
for accessing the EEPROM (due to the shared nature on the i82546).
kern/16373.
- SiS 7012 support from kern/17716 submitted by Quentin Garnier.
- ICH4 support
- Experimental support for nForce MCP, AMD768 and AMD8111.
- Rename the short name for ICH3: "i830M" -> "ICH3"
Always init chip via I/O space.
Map only I/O registers we actually need.
Automatically detect if we need to do bswapping
Add (untested) code to use automagic bswap on CyberPro.
Report video memory amount.
Report if/how bswapping is done.
When initializing a TCP/IP context descriptor, even if a checksum
offload feature is not going to be used, the IPCSS and TUCSS fields
must be initialized.
the brgphy driver; all users of the BCM5400 and BCM5401 need
the DSP patch and the sledgehammer-reset-at-media-set-time.
Also add a DSP patch for the BCM5411 gleaned from Apple's
GMAC driver for Darwin.
Tested with a 3Com 3c996-T (BCM5700 + BCM5401).
* Fix outbound IPv4 header checksums (missing add of an offset).
* When enabling Rx TCP/UDP checksum offload, make sure that IPv4
header checksum Rx offload is enabled on the chip, as well.
* Bump the number of Rx descriptors from 128 to 256.
* Don't use a sliding Tx interrupt window. Instead, just do reap-behind
when we have <= 1/8 of our available descriptors in wm_start().
* Don't use Tx Queue Empty interrupts, and always set the Tx Interrupt
Delay bit in the Tx descriptor.
* In wm_intr(), always call wm_rxintr() and wm_txintr(), regardless of
their respective ISR bits being set. We're here, might as well do some
work.
* Adjust the Tx and Rx interrupt delay timer values. New values from
Intel's driver for FreeBSD via Allen Briggs.
With these changes, NetBSD can sustain > 900Mb/s userland to userland
*without* using TCP checksum offload using Intel PRO/1000 XT cards.
to enable 64-bit data transfers on 64-bit cards when plugged into
a 64-bit slot. Right know the Asante GigaNIX is listed in that
table.
Sigh, there is an EEPROM bit that can be used to detect 64-bit vs
32-bit cards. Unfortunately, at least 2 vendors of 32-bit cards
fail to clear the "DATA64_EN" bit in the EEPROM, which causes the
card to lose badly, because it still manages to detect that it's
plugged into a 64-bit PCI slot. Yay, stupid hardware vendors.
path: Instead of waiting for the if_snd queue to be drained before
giving ownership of the frist descriptor to the chip, do it after
sync'ing all the descriptors for a single packet.
* Get CFG_M64ADDR, CFG_T64ADDR, and CFG_DATA64_EN from the EEPROM.
Note, we still disable CFG_M64ADDR and CFG_T64ADDR later (XXX need
PCI bus capability flags for these).
* Print a message if we're in a 64-bit slot and 64-bit data is
disabled in the EEPROM. Make sure CFG_DATA64_EN is disabled if
we're not in a 64-bit slot.
-treat the builtin font like any other font at runtime
-for that, copy it to malloc()'d memory during attach()
-in early console initialization, if we have to consider a broken card
(VGA_CONSOLE_ATI_BROKEN_FONTSEL), copy the builtin font to another
location in font ram; the attach() code will do the rest
put the "quirk" code into effect again
bit and then read it back in a loop (with appropriate delays) waiting
for it to read back clear.
This fixes a problem where the bus would hang when bringing down
the interface or changing interface flags on a system with a
sufficiently fast CPU (e.g. 2GHz P4 Xeon).
found on many (all?) of PCI-based ATI graphics cards. It is fully optional
and can be enabled by adding `options VGA_CONSOLE_ATI_BROKEN_FONTSEL'
to config file.
- Temporarily remove `quirk' mechanism. Similar code already exists
in pci_quirks.c.
make the number of available font slots variable,
set up a "quirk" mechanism to tell the generic vga code about crippled
VGA adapters which ignore the "fontsel" TS register,
initiate the quirk table with an ATI chip which happened to be on a board
I tested with.
Afaik quite a number of ATI chips suffers from the "loaded fonts don't
work" problem - these should be added.
Bad side effect of my change: The builtin font will be kicked out
always if a VGA_CONSOLE_SCREENTYPE is specified which needs a loaded
font. In early console initialization, we don't know much about the
graphics card, so we have to assume the worst (ie ATI:-).
Written by Bill Paul for FreeBSD. This port started out with the
port done to OpenBSD by nate@openbsd.org, but ended up looking much
more like the port of the ti driver done by drochner@netbsd.org
(they are similar in structure).
int pci_set_powerstate(pci_chipset_tag_t pc, pcitag_t tag, int newstate)
set power state of the device to newstate.
int pci_get_powerstate(pci_chipset_tag_t pc, pcitag_t tag)
get current power state of the device.
In the future, these functions will be used for ACPI support.
to run ABOUT FIRMWARE at the first isp_reset call. We *do* check for
the registers being set with values which would tell us if there's
firmware running or not- but this seems to not always work. It's not
essential, so move on.
Using pdc202xx_setup_channel() for PDC20268 and newer is wrong, and will
cause trap trying to read from a non-existent register on some arches
(e.g. macppc).
pointed out by Makoto Fujiwara on port-macppc.
Reenable statistics interrupts, reading it as 16bit caused it to not be cleared,and thus the eventual interrupt to not be cleared (and this caused an
infinite loop in interrupt routine).
While I'm there count STE_CarrierSenseErrors as oerror (this is incremented
each time we try to send a packet and the link is down).
- when a transmit underrun occurs, only reset the transmit logic. This prevents
the link from going down at the MII level (the same logic can probably
be used for other transmit errors)
- set the urgent threshold to half the start threshold. From experiments
this helps a bit decreasing the number of underruns
Now that underruns don't cause the interface to stop for several seconds,
make back out the default transmit threshold to 128.
- implement SIMPLEQ_REMOVE(head, elm, type, field). whilst it's O(n),
this mirrors the functionality of SLIST_REMOVE() (the other
singly-linked list type) and FreeBSD's STAILQ_REMOVE()
- remove the unnecessary elm arg from SIMPLEQ_REMOVE_HEAD().
this mirrors the functionality of SLIST_REMOVE_HEAD() (the other
singly-linked list type) and FreeBSD's STAILQ_REMOVE_HEAD()
- remove notes about SIMPLEQ not supporting arbitrary element removal
- use SIMPLEQ_FOREACH() instead of home-grown for loops
- use SIMPLEQ_EMPTY() appropriately
- use SIMPLEQ_*() instead of accessing sqh_first,sqh_last,sqe_next directly
- reorder manual page; be consistent about how the types are listed
- other minor cleanups
cause the link to go down for several seconds on my Dlink DFE580-TX (quad-port
DL1002). This behavior is unacceptable, especially as the driver usually
gets several underrun at once, causing the link to be almost unaviable
for several 10s of seconds.
Bump the default transmit threshold to 1504 bytes to avoid transmit underrun.
is down cause the chip to create an IE_UpdateStats condition which can't be
cleared, causing the driver to enter an interrupt loop.
Workaround: mask IE_UpdateStats
- mute the volume for interrupt channel
- add some improvements for device initialization
but recording does not work properly.
Reported and tested by Jan Wagner <waja@cyconet.org>. Thanks.
space address and use it where the mappings of the VGA card are registered
before descenting too deep into "memory" type specific code
(pci_mem_find() gets noisy if it doesn't like the register)
pci-pci bridge (and thus needs its interrupts "swizzled").
Formerly, pci_probe_device() assumed pci busses other than bus zero
were bridged; however, much recent x86 hardware supports multiple
primary pci busses differently -- there is a system-wide bus numbering
scheme. Now, we instead look at the (newly introduced) sc_bridgetag
value in the parent bus to figure out if there's a ppb or equivalent
in the way.
This fixed at least one case where the i386 MP branch gets interrupt
mapping wrong.
NULL for root PCI busses. For busses behind a bridge, it points to
a persistent copy of the bridge's pcitag_t. This can be very useful
for machine-dependent PCI bus enumeration code.
* Implement a machine-dependent pci_enumerate_bus() for sparc64 which
uses OFW device nodes to enumerate the bus. When a PCI bus that is
behind a bridge is attached, pci_attach_hook() allocates a new PCI
chipset tag for the new bus and sets it's "curnode" to the OFW node
of the bridge. This is used as a starting point when enumerating
that bus. Root busses get the OFW node of the host bridge (psycho).
* Garbage-collect "ofpci" and "ofppb" from the sparc64 port.
for a device into two functions:
* pci_probe_device() actually probes/attaches the device specified
by the provide pcitag_t.
* pci_enumerate_bus() enumerates the bus, and calls pci_probe_device()
for each device on the bus. A pci_enumerate_bus_generic() is provided
which implements the old method of doing this: If something found at
dev0/func0, determine number of functions and probe each one.
Machine-dependent code will be able to specify the bus enumeration
routine in the future.
* Garbage-collect some unused stuff.
* Make the Tx window slide along the Tx job queue space, not the
Tx descriptor space. We are more likely to run low on DMA maps
than we are hardware descriptors.
* When forcing an interrupt, make sure that interrupt-delay-enable is
cleared (necessary with last change to make sure it's set for all
descriptors).
* Crank up the Transmit Interrupt Delay Value to 1024 * 1.024 msec. We
really want these to be deferred.
context setup.
* Implement Matt Thomas's sliding Tx interrupt window algorithm,
forcing an interrupt when the Tx desc list is 2/3 consumed.
* Use the Report Packet Sent interrupt, rather than Report Status,
since we use the Tx descriptor to count Tx errors.
errors for narrow transfers after a reselect. FreeBSD sym driver has a comment
about this, but their workaround (disable SCSI gross error reports) doesn't
work for me. Instead dissallow disconnect if the target is not wide
(FreeBSD doens't allow disconnect until the target has been fully probed, which
is why they may not have noticed my problem).
ad