Since commit 7e790dd5fc ("inotify: fix
error paths in inotify_update_watch") inotify changed the manor in which
it gave watch descriptors back to userspace. Previous to this commit
inotify acted like the following:
inotify_add_watch(X, Y, Z) = 1
inotify_rm_watch(X, 1);
inotify_add_watch(X, Y, Z) = 2
but after this patch inotify would return watch descriptors like so:
inotify_add_watch(X, Y, Z) = 1
inotify_rm_watch(X, 1);
inotify_add_watch(X, Y, Z) = 1
which I saw as equivalent to opening an fd where
open(file) = 1;
close(1);
open(file) = 1;
seemed perfectly reasonable. The issue is that quite a bit of userspace
apparently relies on the behavior in which watch descriptors will not be
quickly reused. KDE relies on it, I know some selinux packages rely on
it, and I have heard complaints from other random sources such as debian
bug 558981.
Although the man page implies what we do is ok, we broke userspace so
this patch almost reverts us to the old behavior. It is still slightly
racey and I have patches that would fix that, but they are rather large
and this will fix it for all real world cases. The race is as follows:
- task1 creates a watch and blocks in idr_new_watch() before it updates
the hint.
- task2 creates a watch and updates the hint.
- task1 updates the hint with it's older wd
- task removes the watch created by task2
- task adds a new watch and will reuse the wd originally given to task2
it requires moving some locking around the hint (last_wd) but this should
solve it for the real world and be -stable safe.
As a side effect this patch papers over a bug in the lib/idr code which
is causing a large number WARN's to pop on people's system and many
reports in kerneloops.org. I'm working on the root cause of that idr
bug seperately but this should make inotify immune to that issue.
Signed-off-by: Eric Paris <eparis@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (34 commits)
net: fix build erros with CONFIG_BUG=n, CONFIG_GENERIC_BUG=n
ipv6: skb_dst() can be NULL in ipv6_hop_jumbo().
tg3: Update copyright and driver version
tg3: Disable 5717 serdes and B0 support
tg3: Add reliable serdes detection for 5717 A0
tg3: Fix std rx prod ring handling
tg3: Fix std prod ring nicaddr for 5787 and 57765
sfc: Fix conditions for MDIO self-test
sfc: Fix polling for slow MCDI operations
e1000e: workaround link issues on busy hub in half duplex on 82577/82578
e1000e: MDIO slow mode should always be done for 82577
ixgbe: update copyright dates
ixgbe: Do not attempt to perform interrupts in netpoll when down
cfg80211: fix refcount imbalance when wext is disabled
mac80211: fix queue selection for data frames on monitor interfaces
iwlwifi: silence buffer overflow warning
iwlwifi: disable tx on beacon update notification
iwlwifi: fix iwl_queue_used bug when read_ptr == write_ptr
mac80211: fix endian error
mac80211: add missing sanity checks for action frames
...
Fixed build errors introduced by commit 7ad6848c (ip: fix mc_loop
checks for tunnels with multicast outer addresses)
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes CERT-FI FICORA #341748
Discovered by Olli Jarva and Tuomo Untinen from the CROSS
project at Codenomicon Ltd.
Just like in CVE-2007-4567, we can't rely upon skb_dst() being
non-NULL at this point. We fixed that in commit
e76b2b2567 ("[IPV6]: Do no rely on
skb->dst before it is assigned.")
However commit 483a47d2fe ("ipv6: added
net argument to IP6_INC_STATS_BH") put a new version of the same bug
into this function.
Complicating analysis further, this bug can only trigger when network
namespaces are enabled in the build. When namespaces are turned off,
the dev_net() does not evaluate it's argument, so the dereference
would not occur.
So, for a long time, namespaces couldn't be turned on unless SYSFS was
disabled. Therefore, this code has largely been disabled except by
people turning it on explicitly for namespace development.
With help from Eugene Teo <eugene@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch updates the copyright notice for 2010 and updates the version
number to 3.106.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The B0 revision of the 5717 will not get enough testing by the time
2.6.33 ships. Since the kernel is already at RC3, serdes support
will require too many patches to fix. For these reasons, this patch
disables 5717 serdes support and will refuse to attach to all 5717
devices that are later than an A0 revision.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The serdes status bit does not work as intended for the 5717 A0.
This patch implements an alternative detection scheme that will only be
valid for A0 revisions.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are some tg3 devices that require the driver to post new rx
buffers in smaller increments. Commit
4361935afe, "tg3: Consider
rx_std_prod_idx a hw mailbox" changed how the driver tracks the rx
producer ring updates, but it does not make any special considerations
for the above-mentioned devices. For those devices, it is possible for
the driver to hit the special case path, which updates the hardware
mailbox register but skips updating the shadow software mailbox member.
If the special case path represents the final mailbox update for this
ISR iteration, the hardware and software mailbox values will be out of
sync. Ultimately, this will cause the driver to use a stale mailbox
value on the next iteration, which will appear to the hardware as a
large rx buffer update. Bad things ensue.
The fix is to update the software shadow mailbox member when the special
case path is taken.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reported-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 87668d352a, titled "tg3: Don't
touch RCB nic addresses", tried to avoid assigning the nic address of
the standard producer ring. Unfortunately, the default nic address is
not correct for the 5787, the 5755M, or the 57765. This patch
reenables the old behavior and opts out of the assignment only
for the 5717.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Tested-by: Chow Loong Jin <hyperair@ubuntu.com>
Tested-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
The MDIO self-test should not be run on boards without an MDIO PHY,
such as SFN5122F-R3 and later revisions. It should also not try to
address a specific MMD in an MDIO clause 22 PHY. Check the
mode_support field to decide which mode to use, if any.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the interface is down and we are using polled mode for MCDI
operations, we busy-wait for completion for approximately 1 jiffy
using udelay() and then back off to schedule(). But the completion
will not wake the task, since we are using polled mode! We must use
schedule_timeout_uninterruptible() instead.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes a delay in hardware after every received packet allowing
more time for transmitted packets to go out in between received packets in
half duplex.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A previous 82577 workaround that set the MDIO access speed to slow mode for
every PHY register read/write when the cable is unplugged should instead
set the access mode to always be slow before any PHY register access.
Since the mode bit gets cleared when the PHY is reset, set the mode after
every PHY reset.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch resolves issues seen when running netconsole and rebooting via
reboot -f. The issue was due to the fact that we were attempting to
perform interrupt actions when the q_vectors and rings had already been
freed via the ixgbe_shutdown routines.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'drm-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm: change drm set mode messages as DRM_DEBUG
drm: fix crtc no modes printf + typo
drm/radeon/kms: only evict to GTT if CP is ready
drm/radeon/kms: Fix crash getting TV info with no BIOS.
drm/radeon/kms/rv100: reject modes > 135 Mhz on DVI (v2)
drm/radeon/kms/r6xx+: make irq handler less verbose
drm/radeon/kms: fix up LVDS handling on macs (v2)
Commit ac4c2a3bbe broke the build
of all powerpc boot wrappers.
It attempts to add an include of autoconf.h but used the wrong
path for it. It also adds -D__KERNEL__ to our boot wrapper, both
things that we pretty much didn't do on purpose so far.
We want our boot wrapper to remain independent enough of the kernel
for various reasons, one of them being that you can "wrap" an existing
kernel at distro install time which allows to ship one kernel image
and a set of boot wrappers for different platforms, the wrappers
don't have to be built out of the same kernel build tree.
It's also incorrect to do what the patch does in our boot environment
since we may not have a proper alignment exception handler which means
we may not be able to fixup the few cases where an unaligned access will
need SW emulation (depends on the core variant, could be when crossing
page or segment boundaries for example).
This patch fixes it by putting the old code back in and using the
new "fancy" variant only when CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
is set, which happens not to be set on powerpc since we don't include
autoconf.h. It also reverts the changes to our boot wrapper Makefile.
This means that x86 should, afaik, keep the optimisations since its
boot wrapper does include autoconf.h and define __KERNEL__ (though I
doubt they make that much different outside of slow embedded processors).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
HID: wacom: Add BTN_TOOL_FINGER for pad button reporting
HID: add device IDs for new model of Apple Wireless Keyboard
HID: fix pad button definition in hid-wacom
HID: Support 171 byte variant of Samsung USB IR receiver
HID: blacklist ET&T TC5UH touchscreen controller
If __block_prepare_write() was failed in block_write_begin(), the
allocated blocks can be outside of ->i_size.
But new truncate_pagecache() in vmtuncate() does nothing if new < old.
It means the above usage is not working anymore.
So, this patch fixes it by removing "new < old" check. It would need
more cleanup/change. But, now -rc and truncate working is in progress,
so, this tried to fix it minimum change.
Acked-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's not obvious that copy_from_user() is called with a sane length
parameter here. Even though it currently seems to be correct better
add a check to prevent stack corruption / exploits.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
sys_recvmmsg is reachable via sys_socketcall. So unwire it again since
there is no point in having two entry points for it.
Also put it to the ignore list so we don't get reminded anymore in order
to wire it up.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
...instead of adding a compat ioctl function which would do nothing
as well.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This one will trap, generates shorter code and emits better debug data
than the generic version.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Finally move it to the place where it belongs to and make get rid of
it for !CONFIG_SMP.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This is just a complicated construct which always returns -EINVAL.
Just remove it.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
smp_processor_id() is supposed to work before setup_arch() gets called.
Before that smp_processor_id() may return just an arbitrary value that
is contained in the uninitialized boot lowcore.
So provide the arch function which will override the weak function in
init/main.c.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Make sure compiler won't do weird things with limits. E.g. fetching
them twice may return 2 different values after writable limits are
implemented.
I.e. either use rlimit helpers added in
3e10e716ab
or ACCESS_ONCE if not applicable.
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: linux390@de.ibm.com
Cc: linux-s390@vger.kernel.org
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Add missing compat ptr conversion including two additional
whitespace changes that aren't worth a separate patch.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The TIF_USEDFPU bit is always 0 for s390 and it is not tested anywhere.
Remove the bit. At the same time remove the calls to clear_used_math()
as well. The PF_USED_MATH bit is never set for s390 either.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The code in do_signal sets the TIF_SINGLE_STEP bit and calls
tracehook_signal_handler after the signal frame has been set up.
This causes two SIGTRAP signals to be delivered to the tracer.
Stop setting the TIF_SINGLE_STEP bit in do_signal to get the
correct number of SIGTRAPs.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Clear the TIF_SINGLE_STEP bit in copy_thread. The new process did not get
a PER event of its own. It is wrong deliver a SIGTRAP that was meant for
the parent process.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
If the current task enables / disables PER tracing for itself the
PER control registers need to be loaded in FixPerRegisters.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Setting LC_CTYPE=C breaks localized messages in some setups. With only
LC_COLLATE=C and LC_NUMERIC=C, we get almost all we need, except for not
so defined character classes and tolower()/toupper(). The former is not
a big issue, because we can assume that e.g. [:alpha:] will always
include a-zA-Z and we only ever process ASCII input. The latter seems
only affect arch/sh/tools/gen-mach-types, which we can handle separately.
So after this patch the meaning of ranges like [a-z], the behavior of
sort and join, etc. should be the same everywhere and at the same time
gcc should be able to print localized waring and error messages.
LC_NUMERIC=C might not be necessary, but setting it doesn't hurt.
Reported-by: Simon Horman <horms@verge.net.au>
Reported-by: Sergei Trofimovich <slyfox@inbox.ru>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Tested-by: Simon Horman <horms@verge.net.au>
Tested-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
In an x86 build with CONFIG_KERNEL_LZMA enabled and dash as sh,
arch/x86/boot/compressed/vmlinux.bin.lzma ends with
'\xf0\x7d\x39\x00' (16 bytes) instead of the 4 bytes intended and
the resulting vmlinuz fails to boot. This improves on the
previous behavior, in which the file contained the characters
'-ne ' as well, but not by much.
Previous commits replaced "echo -ne" first with "/bin/echo -ne",
then "printf" in the hope of improving portability, but none of
these commands is guaranteed to support hexadecimal escapes on
POSIX systems. So use the shell to convert from hexadecimal to
octal.
With this change, an LZMA-compressed kernel built with dash as sh
boots correctly again.
Reported-by: Sebastian Dalfuß <sd@sedf.de>
Reported-by: Oliver Hartkopp <oliver@hartkopp.net>
Reported-by: Michael Guntsche <mike@it-loops.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Cc: Michael Tokarev <mjt@tls.msk.ru>
Cc: Alek Du <alek.du@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Without this patch xf86-input-wacom driver wasn't able to properly recognise
pad button events. It was also causing some problems with button mapping.
Signed-off-by: Przemo Firszt <przemo@firszt.eu>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Added device IDs for the new model of the Apple Wireless Keyboard
(November 2009).
Signed-off-by: Christian Schuerer-Waldheim <csw@xray.at>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
This fix is required for xorg driver to recognise 2 pad buttons
Signed-off-by: Przemo Firszt <przemo@firszt.eu>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
In 65f63384 "xen: improve error handling in do_suspend" I said:
- xs_suspend()/xs_resume() and dpm_suspend_noirq()/dpm_resume_noirq() were not
nested in the obvious way.
and changed the ordering of the calls as so:
BEFORE AFTER
xs_suspend dpm_suspend_noirq
dpm_suspend_noirq xs_suspend
*SUSPEND* *SUSPEND*
dpm_resume_noirq dpm_resume_noirq
xs_resume xs_resume
Clearly this is not an improvement and I was talking rubbish.
In particular the new ordering is susceptible to a hang if a xenstore write is
in progress at the point at which the suspend kicks in. When the suspend
process calls xs_suspend it tries to take the request_mutex but if a write is
in progress it could be looping in xenbus_xs.c:read_reply() waiting for
something to arrive on &xs_state.reply_list while holding the request_mutex
(taken in the caller of read_reply).
However if we have done dpm_suspend_noirq before xs_suspend then we won't get
any more xenstore interrupts and process_msg() will never be woken up to add
anything to the reply_list.
Fix this by calling xs_suspend before dpm_suspend_noirq. If dpm_suspend_noirq
fails then make sure we go through the xs_suspend_cancel() code path.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: Stable Kernel <stable@kernel.org>
* korg/drm-radeon-next
drm/radeon/kms: only evict to GTT if CP is ready
drm/radeon/kms: Fix crash getting TV info with no BIOS.
drm/radeon/kms/rv100: reject modes > 135 Mhz on DVI (v2)
drm/radeon/kms/r6xx+: make irq handler less verbose
drm/radeon/kms: fix up LVDS handling on macs (v2)
