Dave Airlie [Sat, 29 Mar 2008 21:51:49 +0000 (07:51 +1000)]
drm/r300: fix bug in r300 userspace hardware wait emission
This interface was originally designed wrong, confusing bit-fields and
integers, major brown paper bag going back many years...
But userspace only ever used 4 values so fix the interface for new
users and fix the implementation to deal with the 4 values userspace
has ever emitted (0x1, 0x2, 0x3, 0x6).
[ conflict in drivers/ide/ide-probe.c fixed manually ]
It turned out that probing order change causes problems for some drives:
http://bugzilla.kernel.org/show_bug.cgi?id=10239
Since root causes are still being investigated and are unlikely to be fixed
before 2.6.25 lets revert this change for now. As a result cable detection
becomes less reliable when compared with 2.6.24 but the affected drives are
useable again.
There have been reported regressions of the SIL 680 driver when using MMIO, so
this makes it only try MMIO on Cell blades where it's known to be necessary
(the host bridge doesn't do PIO on these).
We'll try to find the root problem with MMIO separately.
Mike Frysinger [Fri, 28 Mar 2008 21:41:15 +0000 (14:41 -0700)]
usb net: asix does not really need 10/100mbit
The asix usb driver currently depends on NET_ETHERNET which means you
cannot enable this driver if you only have 1000mbit enabled in your kernel.
Since there is no real dependency between the NET_ETHERNET portion and the
asix driver, simply drop it.
Michael Ellerman [Fri, 28 Mar 2008 01:17:33 +0000 (12:17 +1100)]
Make pasemi_mac.c depend on PPC_PASEMI to prevent link errors
drivers/net/pasemi_mac.c is enabled by CONFIG_PASEMI_MAC, which depends on
PPC64 && PCI. However pasemi_mac.c uses several routines that are only
built when PPC_PASEMI is selected. This can lead to an unbuildable config:
Grant Grundler [Mon, 24 Mar 2008 05:23:10 +0000 (23:23 -0600)]
[netdrvr] tulip_read_eeprom fixes for BUG 4420
If "location" is > "addr_len" bits, the high bits of location would interfere
with the READ_CMD sent to the eeprom controller.
A patch was submitted to bug:
http://bugzilla.kernel.org/show_bug.cgi?id=4420
which simply truncated the "location", read whatever was in "location
modulo addr_len", and returned that value. That avoids confusing the
eeprom but seems like the wrong solution to me.
Correct would be to not read beyond "1 << addr_len" address of the eeprom.
I am submitting two changes to implement this:
1) tulip_read_eeprom will return zero (since we can't return -EINVAL)
if this is attempted (defensive programming).
2) In tulip_core.c, fix the tulip_read_eeprom caller so they don't
iterate past addr_len bits and make sure the entire tp->eeprom[]
array is cleared.
I konw we don't strictly need both. I would prefer both in the tree
since it documents the issue and provides a second "defense" from
the bug from creeping back in.
FUJITA Tomonori [Fri, 28 Mar 2008 22:55:41 +0000 (15:55 -0700)]
sparc64: add the segment boundary checking to IOMMUs while merging SG entries
Some IOMMUs allocate memory areas spanning LLD's segment boundary limit. It
forces low level drivers to have a workaround to adjust scatter lists that the
IOMMU builds. We are in the process of making all the IOMMUs respect the
segment boundary limits to remove such work around in LLDs.
SPARC64 IOMMUs were rewritten to use the IOMMU helper functions and the commit 89c94f2f70d093f59b55d3ea8042d13889169346 made the IOMMUs not allocate memory
areas spanning the segment boundary limit.
However, SPARC64 IOMMUs allocate memory areas first then try to merge them
(while some IOMMUs walk through all the sg entries to see how they can be
merged first and allocate memory areas). So SPARC64 IOMMUs also need the
boundary limit checking when they try to merge sg entries.
Linus Torvalds [Fri, 28 Mar 2008 22:23:01 +0000 (15:23 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
[PATCH] mnt_expire is protected by namespace_sem, no need for vfsmount_lock
[PATCH] do shrink_submounts() for all fs types
[PATCH] sanitize locking in mark_mounts_for_expiry() and shrink_submounts()
[PATCH] count ghost references to vfsmounts
[PATCH] reduce stack footprint in namespace.c
Jesper Juhl [Fri, 28 Mar 2008 21:16:12 +0000 (14:16 -0700)]
driver core: fix small mem leak in driver_add_kobj()
The Coverity checker spotted that we leak the storage allocated to 'name' in
int driver_add_kobj(). The leak looks legit to me - this is the code :
int driver_add_kobj(struct device_driver *drv, struct kobject *kobj,
const char *fmt, ...)
{
va_list args;
char *name;
int ret;
va_start(args, fmt);
name = kvasprintf(GFP_KERNEL, fmt, args);
^^^^^^^^ This dynamically allocates space...
va_end(args);
if (!name)
return -ENOMEM;
return kobject_add(kobj, &drv->p->kobj, "%s", name);
^^^^^^^^ This neglects to free the space allocated
}
Inside kobject_add() a copy of 'name' will be made and used. As far as I can
see, Coverity is correct in flagging this as a leak, but I'd like some
configmation before the patch is applied.
Andrew Morton [Fri, 28 Mar 2008 21:16:09 +0000 (14:16 -0700)]
memstick: suppress uninitialized-var warning
drivers/memstick/host/tifm_ms.c: In function 'tifm_ms_data_event':
drivers/memstick/host/tifm_ms.c:185: warning: 'p_off' may be used uninitialized in this function
Anton Vorontsov [Fri, 28 Mar 2008 21:16:09 +0000 (14:16 -0700)]
mtd: maps/physmap: fix oops in suspend/resume/shutdown ops
# reboot
...
[ 42.351266] Flash device refused suspend due to active operation (state 0)
[ 42.358195] Unable to handle kernel NULL pointer dereference at virtual address 00000078
[ 42.360060] pgd = c7d9c000
[ 42.362769] [00000078] *pgd=a7d8d031, *pte=00000000, *ppte=00000000
[ 42.372902] Internal error: Oops: 17 [#1]
[ 42.376911] Modules linked in:
[ 42.379980] CPU: 0 Not tainted (2.6.25-rc2-10642-ge8f2594-dirty #73)
[ 42.380000] PC is at physmap_flash_shutdown+0x28/0x54
...
[ 42.380000] Backtrace:
[ 42.380000] [<c0130c1c>] (physmap_flash_shutdown+0x0/0x54) from [<c01207c0>] (platform_drv_shutdown+0x20/0x24)
[ 42.380000] r5:28121969 r4:c0229e08
[ 42.380000] [<c01207a0>] (platform_drv_shutdown+0x0/0x24) from [<c011cd40>] (device_shutdown+0x60/0x88)
[ 42.380000] [<c011cce0>] (device_shutdown+0x0/0x88) from [<c003e8a4>] (kernel_restart_prepare+0x2c/0x3c)
[ 42.380000] r4:00000000
[ 42.380000] [<c003e878>] (kernel_restart_prepare+0x0/0x3c) from [<c003ea00>] (kernel_restart+0x14/0x48)
[ 42.380000] [<c003e9ec>] (kernel_restart+0x0/0x48) from [<c003fdc0>] (sys_reboot+0xe8/0x1f8)
[ 42.380000] r4:01234567
[ 42.380000] [<c003fcd8>] (sys_reboot+0x0/0x1f8) from [<c001aa00>] (ret_fast_syscall+0x0/0x2c)
[ 42.380000] r7:00000058 r6:00000004 r5:00000001 r4:00000000
[ 42.380000] Code: 0a000009e7953004e1a00003e1a0e00f (e593f078)
[ 42.650051] ---[ end trace 6d6c26a0fc3141de ]---
Segmentation fault
INIT: no more processes left in this runlevel
While looping for mtd[i]s, we should stop at the mtd[i] == NULL.
This patch also removes unnecessary "if (info)" checks:
suspend/resume/shutdown ops are executed only if probe() is succeeded, so info
is guaranteed to be !NULL.
If write requests need to be split into pieces, the code must not process them
in parallel because the crypto context cannot be shared. So there can be
parallel crypto operations on one part of the write, but only one write bio
can be processed at a time.
This is not optimal and the workqueue code needs to be optimized for parallel
processing, but for now it solves the problem without affecting the
performance of synchronous crypto operation (most of current dm-crypt users).
opening/reading/mmaping BF54x and BF52x framebuffer simultaneously triggers
BUG: failure at mm/nommu.c:470/add_nommu_vma() Add VM_SHARED to the default
vm_flags
The HP Compaq nc6120 has the same PCI sub-device ID as the nx6110, and the
SMBus is used by ACPI for thermal management on the nc6120, so Linux should
not attach a native driver to it. This means that this quirk is unsafe and
has to be removed.
I also added a comment to help developers realize that adding new IDs to this
SMBus unhiding quirk table should be done only with great care, and in
particular only after checking that ACPI is not making use of the SMBus.
Neither of the headers actually compiles when included from userpsace nor
should it be made available as userspace tools should be using the libraries
or at least headers from e2fsprogs.
Alessandro Zummo [Fri, 28 Mar 2008 21:15:59 +0000 (14:15 -0700)]
ixp4xx-beeper: add MODULE_ALIAS
The following patch allows ixp4xx-beeper to be loaded by udev
automatically when compiled as a module with kernel versions 2.4.24 and
greater.
This patch is required because 43cc71eed1250755986da4c0f9898f9a635cb3bf
("platform: prefix MODALIAS with "platform:"") changed the modalias
string to have the extra prefix.
LKG7102D7:~# udevinfo -a -p /sys/devices/platform/ixp4xx-beeper.4
looking at device '/devices/platform/ixp4xx-beeper.4':
KERNEL=="ixp4xx-beeper.4"
SUBSYSTEM=="platform"
DRIVER==""
ATTR{modalias}=="platform:ixp4xx-beeper"
udev therefore tries to modprobe platform:ixp4xx-beeper instead of
ixp4xx-beeper.
LKG7102D7:~# udevtest /sys/devices/platform/ixp4xx-beeper.4
...
import_uevent_var: import into environment: 'PHYSDEVBUS=platform'
import_uevent_var: import into environment: 'MODALIAS=platform:ixp4xx-beeper'
main: looking at device '/devices/platform/ixp4xx-beeper.4' from
subsystem 'platform'
wait_for_sysfs: file '/sys/devices/platform/ixp4xx-beeper.4/bus'
appeared after 0 loops
main: run: 'socket:/org/kernel/udev/monitor'
main: run: '/sbin/modprobe --use-blacklist platform:ixp4xx-beeper'
With this patch, depmod adds an alias line (see below) to
modules.alias which allows modprobe to load the right module.
Andy Whitcroft [Fri, 28 Mar 2008 21:15:58 +0000 (14:15 -0700)]
update checkpatch.pl to version 0.16
This version brings proper quote tracking across lines, and brings the
handling of comments into the same mechanism ensuring nesting is correctly
handled. It brings the usual flurry of fixes for false positives. It also
brings a number of new checks. The most contentious change will likely be
the checks for NR_CPUS as this throws some new warnings in kernel/sched.c.
Of note:
- all new quote tracking across lines
- all new comment tracking
- new more direct, less ambigious wording for some warnings
- recommends mutexes and completions over semaphores
- recommends strict_strto* over simple_strto*
- report on direct use of NR_CPUS
Andy Whitcroft (22):
Version: 0.16
string quote tracking should cross line boundaries
check spacing round -> correctly across newlines
checks for linux/ against asm/ include files should be warnings
standardise on 'required' and 'prohibited'
take the first end of condition when parsing statements
values: cope with unbalanced brackets
preprocessor #elif is not a function
preprocessor #if should not trigger trailing statement checks
test: allow us to limit output to a single error
recommend real mutexes over semaphores
asm checks should mirror those for __asm__
warn on semaphores being used in place of completions
trailing ; on control structure should ignore do {} while ();
recommend strict_strtoX over simple_strtoX
redo comment handling as a quote type
use of NR_CPUS is normally wrong
consistant spacing should only be about spaces
if brace check suppression should only apply to the top-levels
use tr/// to align spacing for operators
move to using four parameter form of substr
check and report modifications to include/asm
Sven Schnelle [Fri, 28 Mar 2008 21:15:55 +0000 (14:15 -0700)]
afs: prevent double cell registration
kafs doesn't check if the cell already exists - so if you do an echo "add
newcell.org 1.2.3.4" >/proc/fs/afs/cells it will try to create this cell
again. kobject will also complain about a double registration. To prevent
such problems, return -EEXIST in that case.
Dmitri Monakhov [Fri, 28 Mar 2008 21:15:52 +0000 (14:15 -0700)]
vfs: fix data leak in nobh_write_end()
Current nobh_write_end() implementation ignore partial writes(copied < len)
case if page was fully mapped and simply mark page as Uptodate, which is
totally wrong because area [pos+copied, pos+len) wasn't updated explicitly in
previous write_begin call. It simply contains garbage from pagecache and
result in data leakage.
As you can see file's page contains garbage from pagecache instead of zeros.
#TEST_CASE_END
Attached patch:
- Add sanity check BUG_ON in order to prevent incorrect usage by caller,
This is function invariant because page can has buffers and in no zero
*fadata pointer at the same time.
- Always attach buffers to page is it is partial write case.
- Always switch back to generic_write_end if page has buffers.
This is reasonable because if page already has buffer then generic_write_begin
was called previously.
drivers/char/drm/ati_pcigart.c: In function 'drm_ati_pcigart_init':
drivers/char/drm/ati_pcigart.c:125: warning: format '%08X' expects type 'unsigned int', but argument 3 has type 'dma_addr_t'
Andrew Morton [Fri, 28 Mar 2008 18:47:34 +0000 (11:47 -0700)]
Avoid false positive warnings in kmap_atomic_prot() with DEBUG_HIGHMEM
I believe http://bugzilla.kernel.org/show_bug.cgi?id=10318 is a false
positive. There's no way in which networking will be using highmem pages
here, so it won't be taking the KM_USER0 kmap slot, so there's no point in
performing these checks.
Cc: Pawel Staszewski <[email protected]> Cc: Ingo Molnar <[email protected]> Acked-by: Christoph Lameter <[email protected]> Cc: "David S. Miller" <[email protected]> Signed-off-by: Andrew Morton <[email protected]>
[ Really sad. We lose almost all real-life coverage of the debug tests
with this patch. Now it will only report problems for the cases where
people actually end up using a HIGHMEM page, not when they just _might_
use one. - Linus ] Signed-off-by: Linus Torvalds <[email protected]>
Roland Dreier [Fri, 28 Mar 2008 17:28:17 +0000 (10:28 -0700)]
RDMA/cxgb3: Program hardware IRD with correct value
Because of a typo in iwch_accept_cr(), the cxgb3 connection handling
code programs the hardware IRD (incoming RDMA read queue depth) with
the value that is passed in for the ORD (outgoing RDMA read queue
depth). In particular this means that if an application passes in IRD
> 0 and ORD = 0 (which is a completely sane and valid thing to do for
an app that expects only incoming RDMA read requests), then the
hardware will end up programmed with IRD = 0 and the app will fail in
a mysterious way.
Fix this by using "ep->ird" instead of "ep->ord" in the intended place.
Ke Wei [Thu, 27 Mar 2008 06:54:23 +0000 (14:54 +0800)]
[SCSI] mvsas : interrupt handling
When a slot is busy, we will not free this slot until slot reset is
completed. When unplugged the disk, we should release all command
tasks with unplugged port that have been sent.
If MVS_USE_TASKLET is defined, we can enable tasklet. Default is off.
Ingo Molnar [Fri, 28 Mar 2008 13:28:03 +0000 (14:28 +0100)]
revert "ACPI: drivers/acpi: elide a non-zero test on a result that is never 0"
Revert commit 1192aeb957402b45f311895f124e4ca41206843c ("ACPI:
drivers/acpi: elide a non-zero test on a result that is never 0")
because it turns out that thermal_cooling_device_register() does
actually return NULL if CONFIG_THERMAL is turned off (then the routine
turns into a dummy inline routine in the header files that returns NULL
unconditionally).
This was found with randconfig testing, causing a crash during bootup:
initcall 0x78878534 ran for 13 msecs: acpi_button_init+0x0/0x51()
Calling initcall 0x78878585: acpi_fan_init+0x0/0x2c()
BUG: unable to handle kernel NULL pointer dereference at 00000000
IP: [<782b8ad0>] acpi_fan_add+0x7d/0xfd
*pde = 00000000
Oops: 0000 [#1]
Modules linked in:
[ARM] 4875/1: Add MODULE_ALIAS to ixp4xx-beeper module
The following patch allows ixp4xx-beeper to be loaded by udev
automatically when compiled as a module with kernel versions 2.4.24
and greater. This patch is required because commit 43cc71eed1250755986da4c0f9898f9a635cb3bf adds "platform:" to the
modalias string.
LKG7102D7:~# udevinfo -a -p /sys/devices/platform/ixp4xx-beeper.4
looking at device '/devices/platform/ixp4xx-beeper.4':
KERNEL=="ixp4xx-beeper.4"
SUBSYSTEM=="platform"
DRIVER==""
ATTR{modalias}=="platform:ixp4xx-beeper"
udev therefore tries to modprobe platform:ixp4xx-beeper instead of
ixp4xx-beeper.
LKG7102D7:~# udevtest /sys/devices/platform/ixp4xx-beeper.4
...
import_uevent_var: import into environment: 'PHYSDEVBUS=platform'
import_uevent_var: import into environment: 'MODALIAS=platform:ixp4xx-beeper'
main: looking at device '/devices/platform/ixp4xx-beeper.4' from
subsystem 'platform'
wait_for_sysfs: file '/sys/devices/platform/ixp4xx-beeper.4/bus'
appeared after 0 loops
main: run: 'socket:/org/kernel/udev/monitor'
main: run: '/sbin/modprobe --use-blacklist platform:ixp4xx-beeper'
With this patch, depmod adds an alias line (see below) to
modules.alias which allows modprobe to load the right module.
modules.alias:
alias platform:ixp4xx-beeper ixp4xx-beeper
Riku Voipio [Fri, 28 Mar 2008 12:08:09 +0000 (13:08 +0100)]
[ARM] 4878/1: Add oabi shim for fstatat64
Ccoreutils and other have started using fstatat64. Thus, we
need a shim for it if we want to support modern oldabi
userlands (such as Debian/arm/lenny) with EABI kernels.
Michael Ellerman [Fri, 28 Mar 2008 08:11:48 +0000 (19:11 +1100)]
[POWERPC] Fix missed hardware breakpoints across multiple threads
There is a bug in the powerpc DABR (data access breakpoint) handling,
which can result in us missing breakpoints if several threads are trying
to break on the same address.
The circumstances are that do_page_fault() calls do_dabr(), this clears
the DABR (sets it to 0) and sets up the signal which will report to
userspace that the DABR was hit. The do_signal() code will restore the DABR
value on the way out to userspace.
If we reschedule before calling do_signal(), __switch_to() will check the
cached DABR value and compare it to the new thread's value, if they match
we don't set the DABR in hardware.
So if two threads have the same DABR value, and we schedule from one to
the other after taking the interrupt for the first thread hitting the DABR,
the second thread will run without the DABR set in hardware.
The cleanest fix is to move the cache update into set_dabr(), that way we
can't forget to do it.
The masking was not at all useless, and it was sensible. We handle
GFP_ZERO in the caller, and passing it down to any page allocator logic
is buggy and wrong.
Patrick McHardy [Fri, 28 Mar 2008 03:28:10 +0000 (20:28 -0700)]
[LLC]: Restrict LLC sockets to root
LLC currently allows users to inject raw frames, including IP packets
encapsulated in SNAP. While Linux doesn't handle IP over SNAP, other
systems do. Restrict LLC sockets to root similar to packet sockets.
[ Modified Patrick's patch to use CAP_NEW_RAW --DaveM ]
Al Viro [Sat, 22 Mar 2008 04:46:23 +0000 (00:46 -0400)]
[PATCH] do shrink_submounts() for all fs types
... and take it out of ->umount_begin() instances. Call with all locks
already taken (by do_umount()) and leave calling release_mounts() to
caller (it will do release_mounts() anyway, so we can just put into
the same list).
Al Viro [Sat, 22 Mar 2008 03:59:49 +0000 (23:59 -0400)]
[PATCH] count ghost references to vfsmounts
make propagate_mount_busy() exclude references from the vfsmounts
that had been isolated by umount_tree() and are just waiting for
release_mounts() to dispose of their ->mnt_parent/->mnt_mountpoint.
Thomas Graf [Thu, 27 Mar 2008 23:08:03 +0000 (16:08 -0700)]
[ESP]: Ensure IV is in linear part of the skb to avoid BUG() due to OOB access
ESP does not account for the IV size when calling pskb_may_pull() to
ensure everything it accesses directly is within the linear part of a
potential fragment. This results in a BUG() being triggered when the
both the IPv4 and IPv6 ESP stack is fed with an skb where the first
fragment ends between the end of the esp header and the end of the IV.
James Bottomley [Wed, 26 Mar 2008 16:26:13 +0000 (09:26 -0700)]
[SCSI] libsas: Warn if ATA device detected but CONFIG_SCSI_SAS_ATA not set
We give a very cryptic error if an ATA device is seen on a SAS port
but libsas isn't compiled to include libata to handle them. Add an
extra warning to explain specifically what the problem is.
James Smart [Fri, 21 Mar 2008 21:18:23 +0000 (17:18 -0400)]
[SCSI] hosts.c: fixes for "no error" reported after error scenarios
This patch corrects some cases in scsi_add_host() that fail, but the "error"
return code was not reset after a prior use which set it to a non-error value.
Jarod Wilson [Tue, 25 Mar 2008 20:47:16 +0000 (16:47 -0400)]
firewire: fw-ohci: plug dma memory leak in AR handler
There's an ugly little memory leak in firewire-ohci's
ar_context_tasklet(), where we're not freeing up some of the memory we
use for each ar_buffer, due to a moving pointer. The problem has been
there for a while, but didn't get noticed until after converting the AR
routines over to use coherent DMA and I started running into I/O stall-
outs with the following message output repeatedly to the console:
PCI-DMA: Out of IOMMU space for 53248 bytes at device 0000:04:09.0
Plugging this leak is definitely necessary, but unfortunately, isn't the
entire answer to my problem, it only increases the amount of I/O that I
can do before hitting the problem. Still working on tracking down the
root cause..
Ivo van Doorn [Thu, 27 Mar 2008 16:15:24 +0000 (17:15 +0100)]
rt2x00: Ignore set_state(STATE_SLEEP) failure
Some hardware never seem to accept the "goto sleep" command, since the legacy
drivers don't have suspend and resume handlers the entire code for it was
basically a educated guess (based on the "enable radio" code).
This patch will only print a warning when the "goto sleep" command fails, and
just continues as usual. Perhaps that means the device will not reach a sleep
state and consumes more power then it should, but it is equally possible it
simply needs some seconds longer to sleep. Anyway, by making the command
non-fatal it will not block the rest of the suspend procedure.
Julia Lawall [Tue, 4 Mar 2008 22:58:59 +0000 (14:58 -0800)]
drivers/net/wireless/iwlwifi/iwl-4965.c: correct use of ! and &
In commit e6bafba5b4765a5a252f1b8d31cbf6d2459da337, a bug was fixed that
involved converting !x & y to !(x & y). The code below shows the same
pattern, and thus should perhaps be fixed in the same way.
This is not tested and clearly changes the semantics, so it is only
something to consider.
The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)
Holger Schurig [Wed, 19 Mar 2008 14:24:21 +0000 (15:24 +0100)]
libertas: fix spinlock recursion bug
This fixes a bug detected by CONFIG_DEBUG_SPINLOCK:
if_cs_get_int_status() is only called from lbs_thread(), via
priv->hw_get_int_status. However, lbs_thread() has already taken the
priv->driver_lock. So it's a fault to take the same lock again here.
Linus Torvalds [Thu, 27 Mar 2008 16:14:07 +0000 (09:14 -0700)]
Merge branch 'avr32-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6
* 'avr32-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6:
avr32: Fix bug in early resource allocation code
avr32: Build fix for CONFIG_BUG=n
avr32: Work around byteswap bug in gcc < 4.2
We need to set up the shared_info pointer once we've mapped the real
shared_info into its fixmap slot. That needs to happen once the general
pagetable setup has been done. Previously, the UP shared_info was set
up one in xen_start_kernel, but that was left pointing to the dummy
shared info. Unfortunately there's no really good place to do a later
setup of the shared_info in UP, so just do it once the pagetable setup
has been done.
xen_irq_enable_direct and xen_sysexit were using "andw $0x00ff,
XEN_vcpu_info_pending(vcpu)" to unmask events and test for pending ones
in one instuction.
Unfortunately, the pending flag must be modified with a locked operation
since it can be set by another CPU, and the unlocked form of this
operation was causing the pending flag to get lost, allowing the processor
to return to usermode with pending events and ultimately deadlock.
The simple fix would be to make it a locked operation, but that's rather
costly and unnecessary. The fix here is to split the mask-clearing and
pending-testing into two instructions; the interrupt window between
them is of no concern because either way pending or new events will
be processed.
This should fix lingering bugs in using direct vcpu structure access too.
The first page of the compound page is determined in follow_huge_addr()
but then PageCompound() only checks if the page is part of a compound page.
PageHead() allows checking if this is indeed the first page of the
compound.
Florian Fainelli [Wed, 26 Mar 2008 21:39:15 +0000 (22:39 +0100)]
rdc321x: GPIO routines bugfixes
This patch fixes the use of GPIO routines which are in the PCI
configuration space of the RDC321x, therefore reading/writing
to this space without spinlock protection can be problematic.
We also now request and free GPIOs and support the MGB100
board, previous code was very AR525W-centric.
projects / linux.git / log