NFS: Fix nfs_wb_page() to always exit with an error or a clean page
It is possible for nfs_wb_page() to sometimes exit with 0 return value, yet
the page is left in a dirty state.
For instance in the case where the server rebooted, and the COMMIT request
failed, then all the previously "clean" pages which were cached by the
server, but were not guaranteed to have been writted out to disk,
have to be redirtied and resent to the server.
The fix is to have nfs_wb_page_priority() check that the page is clean
before it exits...
This fixes a condition that triggers the BUG_ON(PagePrivate(page)) in
nfs_create_request() when we're in the nfs_readpage() path.
Also eliminate a redundant BUG_ON(!PageLocked(page)) while we're at it. It
turns out that clear_page_dirty_for_io() has the exact same test.
SUNRPC: Don't attempt to destroy expired RPCSEC_GSS credentials..
..and always destroy using a 'soft' RPC call. Destroying GSS credentials
isn't mandatory; the server can always cope with a few credentials not
getting destroyed in a timely fashion.
This actually fixes a hang situation. Basically, some servers will decide
that the client is crazy if it tries to destroy an RPC context for which
they have sent an RPCSEC_GSS_CREDPROBLEM, and so will refuse to talk to it
for a while.
The regression therefor probably was introduced by commit 0df7fb74fbb709591301871a38aac7735a1d6583.
The rest of the networking layer uses SOCK_ASYNC_NOSPACE to signal whether
or not we have someone waiting for buffer memory. Convert the SUNRPC layer
to use the same idiom.
Remove the unlikely()s in xs_udp_write_space and xs_tcp_write_space. In
fact, the most common case will be that there is nobody waiting for buffer
space.
SOCK_NOSPACE is there to tell the TCP layer whether or not the cwnd was
limited by the application window. Ensure that we follow the same idiom as
the rest of the networking layer here too.
Finally, ensure that we clear SOCK_ASYNC_NOSPACE once we wake up, so that
write_space() doesn't keep waking things up on xprt->pending.
call_verify() can, under certain circumstances, free the RPC slot. In that
case, our cached pointer 'req = task->tk_rqstp' is invalid. Bug was
introduced in commit 220bcc2afd7011b3e0569fc178331fa983c92c1b (SUNRPC:
Don't call xprt_release in call refresh).
Chuck Lever [Fri, 14 Mar 2008 18:25:53 +0000 (14:25 -0400)]
lockd: document use of mon_id argument in SM_MON requests
Clean up: document the argument type that xdr_encode_common() is
marshalling by introducing a new function. The new function will replace
xdr_encode_common() in just a sec.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever [Fri, 14 Mar 2008 18:25:39 +0000 (14:25 -0400)]
lockd: refactor SM_MON mon_name argument encoder
Clean up: introduce a new XDR encoder specifically for the mon_name
argument of SM_MON requests. This will be updated later to support IPv6
addresses in addition to IPv4 addresses.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever [Fri, 14 Mar 2008 18:18:30 +0000 (14:18 -0400)]
NLM: LOCKD fails to load if CONFIG_SYSCTL is not set
Bruce Fields says:
"By the way, we've got another config-related nit here:
http://bugzilla.linux-nfs.org/show_bug.cgi?id=156
You can build lockd without CONFIG_SYSCTL set, but then the module will
fail to load."
For now, disable the sysctl registration calls in lockd if CONFIG_SYSCTL
is not enabled. This allows the kernel to build properly if PROC_FS or
SYSCTL is not enabled, but an NFS client is desired.
In the long run, we would like to be able to build the kernel with an
NFS client but without lockd. This makes sense, for example, if you want
an NFSv4-only NFS client, as NFSv4 doesn't use NLM at all.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever [Fri, 14 Mar 2008 18:15:18 +0000 (14:15 -0400)]
SUNRPC: Add a default setting for CONFIG_SUNRPC_BIND34
Most distros will want support for rpcbind protocols 3 and 4 to default off
until they have integrated user-space support for the new rpcbind daemon
which supports IPv6 RPC services.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever [Fri, 14 Mar 2008 18:15:11 +0000 (14:15 -0400)]
SUNRPC: Update help Kconfig text
Clean up: refresh the help text for Kconfig items related to the sunrpc
module. Remove obsolete URLs, and make the language consistent among
the options.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever [Fri, 14 Mar 2008 18:14:56 +0000 (14:14 -0400)]
NFS: Always enable NFS direct I/O
Since O_DIRECT is a standard feature that is enabled in most distros,
eliminate the CONFIG_NFS_DIRECTIO build option, and change the
fs/nfs/Makefile to always build in the NFS direct I/O engine.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever [Fri, 14 Mar 2008 18:10:22 +0000 (14:10 -0400)]
NFS: Save the value of the "port=" mount option
During a remount based on the mount options displayed in /proc/mounts, we
want to preserve the original behavior of the mount request. Let's save
the original setting of the "port=" mount option in the mount's nfs_server
structure.
This allows us to simplify the default behavior of port setting for NFSv4
mounts: by default, NFSv2/3 mounts first try an RPC bind to determine the
NFS server's port, unless the user specified the "port=" mount option;
Users can force the client to skip the RPC bind by explicitly specifying
"port=<value>".
NFSv4, by contrast, assumes the NFS server port is 2049 and skips the RPC
bind, unless the user specifies "port=". Users can force an RPC bind for
NFSv4 by explicitly specifying "port=0".
I added a couple of extra comments to clarify this behavior.
Jeff Layton [Fri, 22 Feb 2008 19:50:01 +0000 (14:50 -0500)]
NFS: clean up short packet handling for NFSv4 readdir
Currently, the NFS readdir decoders have a workaround for buggy servers
that send an empty readdir response with the EOF bit unset. If the
server sends a malformed response in some cases, this workaround kicks
in and just returns an empty response rather than returning a proper
error to the caller.
This patch does 3 things:
1) have malformed responses with no entries return error (-EIO)
2) preserve existing workaround for servers that send empty
responses with the EOF marker unset.
3) Add some comments to clarify the logic in decode_readdir().
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Jeff Layton [Fri, 22 Feb 2008 19:50:00 +0000 (14:50 -0500)]
NFS: clean up short packet handling for NFSv3 readdir
Currently, the NFS readdir decoders have a workaround for buggy servers
that send an empty readdir response with the EOF bit unset. If the
server sends a malformed response in some cases, this workaround kicks
in and just returns an empty response rather than returning a proper
error to the caller.
This patch does 3 things:
1) have malformed responses with no entries return error (-EIO)
2) preserve existing workaround for servers that send empty
responses with the EOF marker unset.
3) Add some comments to clarify the logic in nfs3_xdr_readdirres().
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Jeff Layton [Fri, 22 Feb 2008 19:49:59 +0000 (14:49 -0500)]
NFS: clean up short packet handling for NFSv2 readdir
Currently, the NFS readdir decoders have a workaround for buggy servers
that send an empty readdir response with the EOF bit unset. If the
server sends a malformed response in some cases, this workaround kicks
in and just returns an empty response rather than returning a proper
error to the caller.
This patch does 3 things:
1) have malformed responses with no entries return error (-EIO)
2) preserve existing workaround for servers that send empty
responses with the EOF marker unset.
3) Add some comments to clarify the logic in nfs_xdr_readdirres().
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Fred Isaman [Wed, 19 Mar 2008 15:24:39 +0000 (11:24 -0400)]
nfs: don't ignore return value from nfs_pageio_add_request
Ignoring the return value from nfs_pageio_add_request can cause deadlocks.
In read path:
call nfs_pageio_add_request from readpage_async_filler
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_pageio_add_request to return 0, *WITHOUT* adding the original
request.
BUT, since return code is ignored, readpage_async_filler assumes it has
been added, and does nothing further, leaving page locked.
do_generic_mapping_read will eventually call lock_page, resulting in deadlock
In write path:
page is marked dirty by generic_perform_write
nfs_writepages is called
call nfs_pageio_add_request from nfs_page_async_flush
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_page_async_flush to return 0, *WITHOUT* adding the original
request, yet marking the request as locked (PG_BUSY) and in writeback,
clearing dirty marks.
The next time a write is done to the page, deadlock will result as
nfs_write_end calls nfs_update_request
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Linus Torvalds [Wed, 19 Mar 2008 04:34:48 +0000 (21:34 -0700)]
Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
async_tx: avoid the async xor_zero_sum path when src_cnt > device->max_xor
fsldma: Fix the DMA halt when using DMA_INTERRUPT async_tx transfer.
Linus Torvalds [Wed, 19 Mar 2008 04:26:24 +0000 (21:26 -0700)]
IDE: Make taskfile interface more robust wrt unexpected end-of-command
Now that we handle all the special commands using REQ_TYPE_ATA_TASKFILE
rather than using the old REQ_TYPE_ATA_CMD model, we need to also
emulate the lack of full taskfile data that comes with the old command
model (ie when commands are generated with the HDIO_DRIVE_CMD ioctl
rather than using the HDIO_DRIVE_TASK[FILE] ioctls).
In particular, this means that we should handle command completion the
more relaxed way that the old drive_cmd_intr() code did. It allows
commands to finish early even if they don't use up all the data that we
thought we had for them.
This fixes a regression seen by Anders Eriksson where some SMART
commands sent by smartd would cause a boot-time system hang on his
machine because the IDE command handling code didn't realize that the
command had completed.
Ingo Molnar [Sun, 16 Mar 2008 10:14:30 +0000 (11:14 +0100)]
sched: tune multi-core idle balancing
WAKE_IDLE is too agressive on multi-core CPUs with the new
wake-affine code, keep it on for SMT/HT balancing alone
(where there's no cache affinity at all between logical CPUs).
Ingo Molnar [Sat, 15 Mar 2008 16:10:34 +0000 (17:10 +0100)]
sched: wakeup-buddy tasks are cache-hot
Wakeup-buddy tasks are cache-hot - this makes it a bit harder
for the load-balancer to tear them apart. (but it's still possible,
if the load is sufficiently assymetric)
Ingo Molnar [Wed, 19 Mar 2008 00:42:00 +0000 (01:42 +0100)]
sched: improve affine wakeups
improve affine wakeups. Maintain the 'overlap' metric based on CFS's
sum_exec_runtime - which means the amount of time a task executes
after it wakes up some other task.
Use the 'overlap' for the wakeup decisions: if the 'overlap' is short,
it means there's strong workload coupling between this task and the
woken up task. If the 'overlap' is large then the workload is decoupled
and the scheduler will move them to separate CPUs more easily.
( Also slightly move the preempt_check within try_to_wake_up() - this has
no effect on functionality but allows 'early wakeups' (for still-on-rq
tasks) to be correctly accounted as well.)
(the md5's changed because stack slots changed and some registers
get scheduled by gcc in a different order - but otherwise the before
and after assembly is instruction for instruction equivalent.)
Dan Williams [Wed, 19 Mar 2008 04:23:59 +0000 (21:23 -0700)]
async_tx: avoid the async xor_zero_sum path when src_cnt > device->max_xor
If the channel cannot perform the operation in one call to
->device_prep_dma_zero_sum, then fallback to the xor+page_is_zero path.
This only affects users with arrays larger than 16 devices on iop13xx or
32 devices on iop3xx.
Cc: <stable@kernel.org> Cc: Neil Brown <neilb@suse.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Zhang Wei [Wed, 19 Mar 2008 01:45:00 +0000 (18:45 -0700)]
fsldma: Fix the DMA halt when using DMA_INTERRUPT async_tx transfer.
The DMA_INTERRUPT async_tx is a NULL transfer, thus the BCR(count register)
is 0. When the transfer started with a byte count of zero, the DMA
controller will triger a PE(programming error) event and halt, not a normal
interrupt. I add special codes for PE event and DMA_INTERRUPT
async_tx testing.
Signed-off-by: Zhang Wei <wei.zhang@freescale.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Linus Torvalds [Tue, 18 Mar 2008 14:49:59 +0000 (07:49 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
HID: update key codes for Apple aluminium
HID: fix comment in hid_input_report()
HID: BADPAD entry for NATSU Playstation USB adapter
HID: Use DIV_ROUND_UP
HID: remove HID_QUIRK_APPLE_ISO_KEYBOARD for 4th generation macbook
Linus Torvalds [Tue, 18 Mar 2008 14:43:14 +0000 (07:43 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
Revert "unexport bio_{,un}map_user"
relay: fix subbuf_splice_actor() adding too many pages
The ps2esdi driver was marked as BROKEN more than two years ago due to being
Linus Torvalds [Tue, 18 Mar 2008 14:32:23 +0000 (07:32 -0700)]
Merge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6
* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm/ati_pcigart: fix the PCIGART to use drm_pci to allocate GART table.
drm/radeon: fixup RV550 chip family
drm/via: attempt again to stabilise the AGP DMA command submission.
drm: Fix race that can lockup the kernel
F5 and F6 have no second function printed on them. Thus their definitions have
been removed from the table.
KEY_CYCLEWINDOWS doesn't name the function of Mac OS X' Expose properly and
because we couldn't find a better key code, we decided to use KEY_FN_F4
instead.
We also changed KEY_BACK and KEY_FORWARD, which apply to browser functions, to
KEY_PREVIOUSSONG and KEY_NEXTSONG, since the keys are intended to control a
music player.
Signed-off-by: Michael Hanselmann <linux-kernel@hansmi.ch> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Holger Macht [Wed, 12 Mar 2008 00:07:27 +0000 (01:07 +0100)]
ACPI: Set flag DOCK_UNDOCKING when triggered via sysfs
begin_undock() is only called when triggered via a acpi notify handler
(pressing the undock button on the dock station), but complete_undock() is
always called after the eject. So if a undock is triggered through a sysfs
write, the flag DOCK_UNDOCKING has to be set for the dock station,
too. Otherwise this will freeze the system hard.
Signed-off-by: Holger Macht <hmacht@suse.de> Acked-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Len Brown <len.brown@intel.com>
Julia Lawall [Tue, 4 Mar 2008 23:00:13 +0000 (15:00 -0800)]
asus_acpi: remove misleading mask
led_out is boolean, so there is no functional change here,
but apparently an extra mask with 1 caused some style checkers
to flag this as logic bug.
Signed-off-by: Julia Lawall <julia@diku.dk> Acked-by: Luca Tettamanti <kronos.it@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Len Brown <len.brown@intel.com>
ACPI: battery: Don't return -EFAIL on broken packages.
Acer BIOS has a bug which is exposed when a dead battery is present.
The package template that is used to describe battery status is
over-written with sane values when the battery is live.
But when the batter is dead, a bogus reference in the template
is used. In this case, Linux returns a fault, when instead
it should simply return that it doesn't know the missing value.
Mark Lord [Mon, 17 Mar 2008 20:04:23 +0000 (16:04 -0400)]
pciehp: don't enable slot unless forced
This fixes a 2.6.25 regression reported by Alex Chiang.
Invoke pciehp_enable_slot() at startup only when pciehp_force=1.
Some HP equipment apparently cannot cope with it otherwise.
This restores the (previously working) 2.6.24 behaviour here,
while allowing machines that need a kick to use pciehp_force=1.
This was the original design back in October 2007,
but Kristen suggested we try without it first:
Kristen Carlson Accardi wrote:
>I think it would be ok to try allowing the slot to be enabled when not
>using pciehp_force mode. We can wrap it later if it proves to break things
This ended up breaking one of Alex's setups,
so it's time to put the wrapper back in now.
Signed-off-by: Mark Lord <mlord@pobox.com> Acked-by: Alex Chiang <achiang@hp.com> Acked-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Jens Axboe [Mon, 17 Mar 2008 20:14:40 +0000 (21:14 +0100)]
Revert "unexport bio_{,un}map_user"
Outside users like asmlib uses the mapping functions. API wise, the
export is definitely sane. It's a better idea to keep this export
than to require external users to open-code this piece of code instead.
slub page alloc fallback: Enable interrupts for GFP_WAIT.
The fallback path needs to enable interrupts like done for
the other page allocator calls. This was not necessary with
the alternate fast path since we handled irq enable/disable in
the slow path. The regular fastpath handles irq enable/disable
around calls to the slow path so we need to restore the proper
status before calling the page allocator from the slowpath.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Linus Torvalds [Mon, 17 Mar 2008 16:52:24 +0000 (09:52 -0700)]
Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
ahci: Add Marvell 6121 SATA support
pata_ali: use atapi_cmd_type() to determine cmd type instead of transfer size
ahci: implement skip_host_reset parameter
ahci: request all PCI BARs
devres: implement pcim_iomap_regions_request_all()
libata-acpi: improve dock event handling
Tejun Heo [Tue, 11 Mar 2008 02:35:00 +0000 (11:35 +0900)]
pata_ali: use atapi_cmd_type() to determine cmd type instead of transfer size
pata_ali was using qc->nbytes to determine whether a command is
data transfer type or not. As now qc->nbytes can be extended by
padding and draining buffers, these tests are not useful anymore.
Use atapi_cmd_type() instead.
Signed-off-by: Tejun Heo <htejun@gmail.com> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Tejun Heo [Mon, 10 Mar 2008 01:25:25 +0000 (10:25 +0900)]
ahci: implement skip_host_reset parameter
Under certain circumstances (SSP turned off by the BIOS) and for
debugging purposes, skipping global controller reset is helpful. Add
a kernel parameter for it.
Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Tejun Heo [Tue, 11 Mar 2008 10:52:31 +0000 (19:52 +0900)]
ahci: request all PCI BARs
ahci is often implemented with accompanying SFF compatible interface
and legacy IDE driver may attach to the legacy IO ports when the
controller is already claimed by ahci and vice-versa. This patch
makes ahci use pcim_iomap_regions_request_all() so that all IO regions
are claimed on attach.
Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Some drivers need to reserve all PCI BARs to prevent other drivers
misusing unoccupied BARs. pcim_iomap_regions_request_all() requests
all BARs and iomap specified BARs.
Signed-off-by: Tejun Heo <htejun@gmail.com> Cc: Greg Kroah-Hartman <gregkh@suse.de> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: Jeff Garzik <jeff@garzik.org> Signed-off-by: Jeff Garzik <jeff@garzik.org>
The problem can be triggered with a high amount of host->guest traffic.
I think its the following race:
poll says netif_rx_complete
poll calls enable_cb
enable_cb opens the interrupt mask
a new packet comes, an interrupt is triggered----\
enable_cb sees that there is more work |
enable_cb disables the interrupt |
. V
. interrupt is delivered
. skb_recv_done does atomic napi test, ok
some waiting disable_cb is called->check fails->bang!
.
poll would do napi check
poll would do disable_cb
The fix is to let enable_cb not disable the interrupt again, but expect the
caller to do the cleanup if it returns false. In that case, the interrupt is
only disabled, if the napi test_set_bit was successful.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (cleaned up doco)
Rusty Russell [Tue, 18 Mar 2008 03:58:15 +0000 (22:58 -0500)]
virtio: handle > 2 billion page balloon targets
If the host asks for a huge target towards_target() can overflow, and
we up oops as we try to release more pages than we have. The simple
fix is to use a 64-bit value.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Anthony Liguori [Sun, 2 Mar 2008 22:37:48 +0000 (16:37 -0600)]
virtio: Use spin_lock_irqsave/restore for virtio-pci
virtio-pci acquires its spin lock in an interrupt context so it's necessary
to use spin_lock_irqsave/restore variants. This patch fixes guest SMP when
using virtio devices in KVM.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Thomas Hellstrom [Mon, 17 Mar 2008 00:07:20 +0000 (10:07 +1000)]
drm/via: attempt again to stabilise the AGP DMA command submission.
It's worth remembering that all new bright ideas on how to make this command reader work properly and according to docs will probably fail :( Bring in some old code.
Also allow a larger SG-DMA download stride, and remove unnecessary waits for
command regulators pauses.
Mike Isely [Thu, 13 Mar 2008 20:30:35 +0000 (15:30 -0500)]
drm: Fix race that can lockup the kernel
The i915_vblank_swap() function schedules an automatic buffer swap
upon receipt of the vertical sync interrupt. Such an operation is
lengthy so it can't be allowed to happen in normal interrupt context,
thus the DRM implements this by scheduling the work in a kernel
softirq-scheduled tasklet. In order for the buffer swap to work
safely, the DRM's central lock must be taken, via a call to
drm_lock_take() located in drivers/char/drm/drm_irq.c within the
function drm_locked_tasklet_func(). The lock-taking logic uses a
non-interrupt-blocking spinlock to implement the manipulations needed
to take the lock. This semantic would be safe if all attempts to use
the spinlock only happen from process context. However this buffer
swap happens from softirq context which is really a form of interrupt
context. Thus we have an unsafe situation, in that
drm_locked_tasklet_func() can block on a spinlock already taken by a
thread in process context which will never get scheduled again because
of the blocked softirq tasklet. This wedges the kernel hard.
To trigger this bug, run a dual-head cloned mode configuration which
uses the i915 drm, then execute an opengl application which
synchronizes buffer swaps against the vertical sync interrupt. In my
testing, a lockup always results after running anywhere from 5 minutes
to an hour and a half. I believe dual-head is needed to really
trigger the problem because then the vertical sync interrupt handling
is no longer predictable (due to being interrupt-sourced from two
different heads running at different speeds). This raises the
probability of the tasklet trying to run while the userspace DRI is
doing things to the GPU (and manipulating the DRM lock).
The fix is to change the relevant spinlock semantics to be the
interrupt-blocking form. After this change I am no longer able to
trigger the lockup; the longest test run so far was 20 hours (test
stopped after that point).
Note: I have examined the places where this spinlock is being
employed; all are reasonably short bounded sequences and should be
suitable for interrupts being blocked without impacting overall kernel
interrupt response latency.
Signed-off-by: Mike Isely <isely@pobox.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
Helge Deller [Wed, 26 Dec 2007 17:07:01 +0000 (18:07 +0100)]
[PARISC] head.S: section mismatch fixes
- move boot_args[] into the init section
- move $global$ into the read_mostly section
- fix the following two section mismatches:
WARNING: vmlinux.o(.text+0x9c): Section mismatch: reference to .init.text:start_kernel (between '$pgt_fill_loop' and '$is_pa20')
WARNING: vmlinux.o(.text+0xa0): Section mismatch: reference to .init.text:start_kernel (between '$pgt_fill_loop' and '$is_pa20')
projects / linux-2.6 / log