]> git.karo-electronics.de Git - karo-tx-linux.git/log
karo-tx-linux.git
16 years agouml: missed kmalloc() in pcap_user.c
Al Viro [Tue, 5 Aug 2008 20:14:05 +0000 (16:14 -0400)]
uml: missed kmalloc() in pcap_user.c

commit 296cd66f7f6e130fe08e6880ecb13c3fc615a8db upstream

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agouml: track and make up lost ticks
Jeff Dike [Tue, 5 Aug 2008 20:14:05 +0000 (16:14 -0400)]
uml: track and make up lost ticks

commit fe2cc53ee013a4d4d0317d418e7019fe6533a5a8 upstream

Alarm delivery could be noticably late in the !CONFIG_NOHZ case because lost
ticks weren't being taken into account.  This is now treated more carefully,
with the time between ticks being calculated and the appropriate number of
ticks delivered to the timekeeping system.

Cc: Nix <nix@esperi.org.uk>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agouml: physical memory shouldn't include initial stack
Jeff Dike [Tue, 5 Aug 2008 20:14:04 +0000 (16:14 -0400)]
uml: physical memory shouldn't include initial stack

commit 60a2988aea701a6424809a5432bf068667aac177 upstream

The top of physical memory should be below the initial process stack, not the
top of the address space, at least for as long as the stack isn't known to the
kernel VM system and appropriately reserved.

Cc: "Christopher S. Aker" <caker@theshore.net>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agouml: fix bad NTP interaction with clock
Jeff Dike [Tue, 5 Aug 2008 20:14:03 +0000 (16:14 -0400)]
uml: fix bad NTP interaction with clock

commit cfd28f6695d0fc047478480791a21bdd4967f98e upstream

UML's supposed nanosecond clock interacts badly with NTP when NTP
decides that the clock has drifted ahead and needs to be slowed down.
Slowing down the clock is done by decrementing the cycle-to-nanosecond
multiplier, which is 1.  Decrementing that gives you 0 and time is
stopped.

This is fixed by switching to a microsecond clock, with a multiplier
of 1000.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agouml: fix build when SLOB is enabled
Jeff Dike [Tue, 5 Aug 2008 20:14:03 +0000 (16:14 -0400)]
uml: fix build when SLOB is enabled

commit 43f5b3085fdd27c4edf535d938b2cb0ccead4f75 upstream

Reintroduce uml_kmalloc for the benefit of UML libc code.  The
previous tactic of declaring __kmalloc so it could be called directly
from the libc side of the house turned out to be getting too intimate
with slab, and it doesn't work with slob.

So, the uml_kmalloc wrapper is back.  It calls kmalloc or whatever
that translates into, and libc code calls it.

kfree is left alone since that still works, leaving a somewhat
inconsistent API.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agosparc64: Do not clobber %g7 in setcontext() trap.
David S. Miller [Fri, 1 Aug 2008 03:40:46 +0000 (20:40 -0700)]
sparc64: Do not clobber %g7 in setcontext() trap.

[ Upstream commit 0a4949c4414af2eb91414bcd8e2a8ac3706f7dde ]

That's the userland thread register, so we should never try to change
it like this.

Based upon glibc bug nptl/6577 and suggestions by Jakub Jelinek.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agosparc64: FUTEX_OP_ANDN fix
Mikael Pettersson [Wed, 30 Jul 2008 22:40:50 +0000 (15:40 -0700)]
sparc64: FUTEX_OP_ANDN fix

[ Upstream commit d72609e17fd93bb2f7e0f7e1bdc70b6d20e43843 ]

Correct sparc64's implementation of FUTEX_OP_ANDN to do a
bitwise negate of the oparg parameter before applying the
AND operation. All other archs that support FUTEX_OP_ANDN
either negate oparg explicitly (frv, ia64, mips, sh, x86),
or do so indirectly by using an and-not instruction (powerpc).
Since sparc64 has and-not, I chose to use that solution.

I've not found any use of FUTEX_OP_ANDN in glibc so the
impact of this bug is probably minor. But other user-space
components may try to use it so it should still get fixed.

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoipv6: Fix ip6_xmit to send fragments if ipfragok is true
Wei Yongjun [Fri, 1 Aug 2008 03:46:47 +0000 (20:46 -0700)]
ipv6: Fix ip6_xmit to send fragments if ipfragok is true

[ Upstream commit 77e2f14f71d68d05945f1d30ca55b5194d6ab1ce ]

SCTP used ip6_xmit() to send fragments after received ICMP packet too
big message. But while send packet used ip6_xmit, the skb->local_df is
not initialized. So when skb if enter ip6_fragment(), the following
code will discard the skb.

ip6_fragment(...)
{
    if (!skb->local_df) {
        ...
        return -EMSGSIZE;
    }
    ...
}

SCTP do the following step:
1. send packet ip6_xmit(skb, ipfragok=0)
2. received ICMP packet too big message
3. if PMTUD_ENABLE: ip6_xmit(skb, ipfragok=1)

This patch fixed the problem by set local_df if ipfragok is true.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agorandom32: seeding improvement
Stephen Hemminger [Wed, 30 Jul 2008 23:29:19 +0000 (16:29 -0700)]
random32: seeding improvement

[ Upstream commit 697f8d0348a652593d195a13dd1067d9df911a82 ]

The rationale is:
   * use u32 consistently
   * no need to do LCG on values from (better) get_random_bytes
   * use more data from get_random_bytes for secondary seeding
   * don't reduce state space on srandom32()
   * enforce state variable initialization restrictions

Note: the second paper has a version of random32() with even longer period
and a version of random64() if needed.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agodccp: change L/R must have at least one byte in the dccpsf_val field
Arnaldo Carvalho de Melo [Wed, 13 Aug 2008 20:48:39 +0000 (13:48 -0700)]
dccp: change L/R must have at least one byte in the dccpsf_val field

commit 3e8a0a559c66ee9e7468195691a56fefc3589740 upstream

Thanks to Eugene Teo for reporting this problem.

Signed-off-by: Eugene Teo <eugeneteo@kernel.sg>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoacer-wmi: Fix wireless and bluetooth on early AMW0 v2 laptops
Carlos Corbacho [Fri, 15 Aug 2008 16:30:03 +0000 (16:30 +0000)]
acer-wmi: Fix wireless and bluetooth on early AMW0 v2 laptops

commit 5c742b45dd5fbbb6cf74d3378341704f4b23c5e8 upstream

In the old acer_acpi, I discovered that on some of the newer AMW0 laptops
that supported the WMID methods, they don't work properly for setting the
wireless and bluetooth values.

So for the AMW0 V2 laptops, we want to use both the 'old' AMW0 and the
'new' WMID methods for setting wireless & bluetooth to guarantee we always
enable it.

This was fixed in acer_acpi some time ago, but I forgot to port the patch
over to acer-wmi when it was merged.

(Without this patch, early AMW0 V2 laptops such as the Aspire 5040 won't
work with acer-wmi, where-as they did with the old acer_acpi).

AK: fix compilation

Signed-off-by: Carlos Corbacho <carlos@strangeworlds.co.uk>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoCIFS: if get root inode fails during mount, cleanup tree connection
Steve French [Fri, 15 Aug 2008 19:05:06 +0000 (19:05 +0000)]
CIFS: if get root inode fails during mount, cleanup tree connection

commit 2c731afb0d4ba16018b400c75665fbdb8feb2175 upstream

Signed-off-by: Steve French <sfrench@us.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoCIFS: mount of IPC$ breaks with iget patch
Steve French [Fri, 15 Aug 2008 19:05:03 +0000 (19:05 +0000)]
CIFS: mount of IPC$ breaks with iget patch

commit ad661334b8ae421154b121ee6ad3b56807adbf11 upstream

In looking at network named pipe support on cifs, I noticed that
Dave Howell's iget patch:

    iget: stop CIFS from using iget() and read_inode()

broke mounts to IPC$ (the interprocess communication share), and don't
handle the error case (when getting info on the root inode fails).

Thanks to Gunter who noted a typo in a debug line in the original
version of this patch.

CC: David Howells <dhowells@redhat.com>
CC: Gunter Kukkukk <linux@kukkukk.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoide-cd: fix endianity for the error message in cdrom_read_capacity
Petr Tesarik [Wed, 6 Aug 2008 00:05:07 +0000 (00:05 +0000)]
ide-cd: fix endianity for the error message in cdrom_read_capacity

commit 938bb03d188a1e688fb0bcae49788f540193e80a uptream

Aesthetic regards aside, commit e8e7b9eb11c34ee18bde8b7011af41938d1ad667
still leaves a bug in the error message, because it uses the unconverted
big-endian value for printk.

Fix this by using a local variable in machine byte order. The result is
correct, more readable, and also produces slightly shorter code on i386.

Signed-off-by: Petr Tesarik <ptesarik@suse.cz>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: <stable@kernel.org>
Acked-by: Borislav Petkov <petkovbb@gmail.com>
[bart: __u32 -> u32]
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoipvs: Fix possible deadlock in estimator code
Sven Wegener [Thu, 14 Aug 2008 03:55:22 +0000 (03:55 +0000)]
ipvs: Fix possible deadlock in estimator code

commit 8ab19ea36c5c5340ff598e4d15fc084eb65671dc upstream

There is a slight chance for a deadlock in the estimator code. We can't call
del_timer_sync() while holding our lock, as the timer might be active and
spinning for the lock on another cpu. Work around this issue by using
try_to_del_timer_sync() and releasing the lock. We could actually delete the
timer outside of our lock, as the add and kill functions are only every called
from userspace via [gs]etsockopt() and are serialized by a mutex, but better
make this explicit.

Signed-off-by: Sven Wegener <sven.wegener@stealer.net>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agomatrox maven: fix a broken error path
Jean Delvare [Tue, 12 Aug 2008 23:20:04 +0000 (23:20 +0000)]
matrox maven: fix a broken error path

commit 5ede40f87957c6ededf9284c8339722a97b9dfb6 upstream

I broke an error path with d03c21ec0be7787ff6b75dcf56c0e96209ccbfbd,
sorry about that.

The machine will crash if the i2c_attach_client() or maven_init_client()
calls fail, although nobody has yet reported this happening.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Petr Vandrovec <VANDROVE@vc.cvut.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agomlock() fix return values
KOSAKI Motohiro [Tue, 5 Aug 2008 00:20:05 +0000 (00:20 +0000)]
mlock() fix return values

commit a477097d9c37c1cf289c7f0257dffcfa42d50197 upstream

Halesh says:

Please find the below testcase provide to test mlock.

Test Case :
===========================

#include <sys/resource.h>
#include <stdio.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <unistd.h>
#include <sys/mman.h>
#include <fcntl.h>
#include <errno.h>
#include <stdlib.h>

int main(void)
{
  int fd,ret, i = 0;
  char *addr, *addr1 = NULL;
  unsigned int page_size;
  struct rlimit rlim;

  if (0 != geteuid())
  {
   printf("Execute this pgm as root\n");
   exit(1);
  }

  /* create a file */
  if ((fd = open("mmap_test.c",O_RDWR|O_CREAT,0755)) == -1)
  {
   printf("cant create test file\n");
   exit(1);
  }

  page_size = sysconf(_SC_PAGE_SIZE);

  /* set the MEMLOCK limit */
  rlim.rlim_cur = 2000;
  rlim.rlim_max = 2000;

  if ((ret = setrlimit(RLIMIT_MEMLOCK,&rlim)) != 0)
  {
   printf("Cant change limit values\n");
   exit(1);
  }

  addr = 0;
  while (1)
  {
  /* map a page into memory each time*/
  if ((addr = (char *) mmap(addr,page_size, PROT_READ |
PROT_WRITE,MAP_SHARED,fd,0)) == MAP_FAILED)
  {
   printf("cant do mmap on file\n");
   exit(1);
  }

  if (0 == i)
    addr1 = addr;
  i++;
  errno = 0;
  /* lock the mapped memory pagewise*/
  if ((ret = mlock((char *)addr, 1500)) == -1)
  {
   printf("errno value is %d\n", errno);
   printf("cant lock maped region\n");
   exit(1);
  }
  addr = addr + page_size;
 }
}
======================================================

This testcase results in an mlock() failure with errno 14 that is EFAULT,
but it has nowhere been specified that mlock() will return EFAULT.  When I
tested the same on older kernels like 2.6.18, I got the correct result i.e
errno 12 (ENOMEM).

I think in source code mlock(2), setting errno ENOMEM has been missed in
do_mlock() , on mlock_fixup() failure.

SUSv3 requires the following behavior frmo mlock(2).

[ENOMEM]
    Some or all of the address range specified by the addr and
    len arguments does not correspond to valid mapped pages
    in the address space of the process.

[EAGAIN]
    Some or all of the memory identified by the operation could not
    be locked when the call was made.

This rule isn't so nice and slighly strange.  but many people think
POSIX/SUS compliance is important.

Reported-by: Halesh Sadashiv <halesh.sadashiv@ap.sony.com>
Tested-by: Halesh Sadashiv <halesh.sadashiv@ap.sony.com>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoposix-timers: do_schedule_next_timer: fix the setting of ->si_overrun
Oleg Nesterov [Tue, 12 Aug 2008 15:30:09 +0000 (15:30 +0000)]
posix-timers: do_schedule_next_timer: fix the setting of ->si_overrun

commit 54da1174922cddd4be83d5a364b2e0fdd693f513 upstream

do_schedule_next_timer() sets info->si_overrun = timr->it_overrun_last,
this discards the already accumulated overruns.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Mark McLoughlin <markmc@redhat.com>
Cc: Oliver Pinter <oliver.pntr@gmail.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoposix-timers: fix posix_timer_event() vs dequeue_signal() race
Oleg Nesterov [Tue, 12 Aug 2008 15:30:06 +0000 (15:30 +0000)]
posix-timers: fix posix_timer_event() vs dequeue_signal() race

commit ba661292a2bc6ddd305a212b0526e5dc22195fe7 upstream

The bug was reported and analysed by Mark McLoughlin <markmc@redhat.com>,
the patch is based on his and Roland's suggestions.

posix_timer_event() always rewrites the pre-allocated siginfo before sending
the signal. Most of the written info is the same all the time, but memset(0)
is very wrong. If ->sigq is queued we can race with collect_signal() which
can fail to find this siginfo looking at .si_signo, or copy_siginfo() can
copy the wrong .si_code/si_tid/etc.

In short, sys_timer_settime() can in fact stop the active timer, or the user
can receive the siginfo with the wrong .si_xxx values.

Move "memset(->info, 0)" from posix_timer_event() to alloc_posix_timer(),
change send_sigqueue() to set .si_overrun = 0 when ->sigq is not queued.
It would be nice to move the whole sigq->info initialization from send to
create path, but this is not easy to do without uglifying timer_create()
further.

As Roland rightly pointed out, we need more cleanups/fixes here, see the
"FIXME" comment in the patch. Hopefully this patch makes sense anyway, and
it can mask the most bad implications.

Reported-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Mark McLoughlin <markmc@redhat.com>
Cc: Oliver Pinter <oliver.pntr@gmail.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoradeonfb: fix accel engine hangs
David Miller [Tue, 12 Aug 2008 23:20:06 +0000 (23:20 +0000)]
radeonfb: fix accel engine hangs

commit 969830b2fedf8336c41d6195f49d250b1e166ff8 upstream

Some chips appear to have the 2D engine hang during screen redraw,
typically in a sequence of copyarea operations. This appear to be
solved by adding a flush of the engine destination pixel cache
and waiting for the engine to be idle before issuing the accel
operation. The performance impact seems to be fairly small.

Here is a trace on an RV370 (PCI device ID 0x5b64), it records the
RBBM_STATUS register, then the source x/y, destination x/y, and
width/height used for the copy:

----------------------------------------
radeonfb_prim_copyarea: STATUS[00000140] src[210:70] dst[210:60] wh[a0:10]
radeonfb_prim_copyarea: STATUS[00000140] src[2b8:70] dst[2b8:60] wh[88:10]
radeonfb_prim_copyarea: STATUS[00000140] src[348:70] dst[348:60] wh[40:10]
radeonfb_prim_copyarea: STATUS[80020140] src[390:70] dst[390:60] wh[88:10]
radeonfb_prim_copyarea: STATUS[8002613f] src[40:80] dst[40:70] wh[28:10]
radeonfb_prim_copyarea: STATUS[80026139] src[a8:80] dst[a8:70] wh[38:10]
radeonfb_prim_copyarea: STATUS[80026133] src[e8:80] dst[e8:70] wh[80:10]
radeonfb_prim_copyarea: STATUS[8002612d] src[170:80] dst[170:70] wh[30:10]
radeonfb_prim_copyarea: STATUS[80026127] src[1a8:80] dst[1a8:70] wh[8:10]
radeonfb_prim_copyarea: STATUS[80026121] src[1b8:80] dst[1b8:70] wh[88:10]
radeonfb_prim_copyarea: STATUS[8002611b] src[248:80] dst[248:70] wh[68:10]
----------------------------------------

When things are going fine the copies complete before the next ROP is
even issued, but all of a sudden the 2D unit becomes active (bit 17 in
RBBM_STATUS) and the FIFO retry (bit 13) and FIFO pipeline busy (bit
14) are set as well.  The FIFO begins to backup until it becomes full.

What happens next is the radeon_fifo_wait() times out, and we access
the chip illegally leading to a bus error which usually wedges the
box.  None of this makes it to the console screen, of course :-)
radeon_fifo_wait() should be modified to reset the accelerator when
this timeout happens instead of programming the chip anyways.

----------------------------------------
radeonfb: FIFO Timeout !
ERROR(0): Cheetah error trap taken afsr[0010080005000000] afar[000007f900800e40] TL1(0)
ERROR(0): TPC[595114] TNPC[595118] O7[459788] TSTATE[11009601]
ERROR(0): TPC<radeonfb_copyarea+0xfc/0x248>
ERROR(0): M_SYND(0),  E_SYND(0), Privileged
ERROR(0): Highest priority error (0000080000000000) "Bus error response from system bus"
ERROR(0): D-cache idx[0] tag[0000000000000000] utag[0000000000000000] stag[0000000000000000]
ERROR(0): D-cache data0[0000000000000000] data1[0000000000000000] data2[0000000000000000] data3[0000000000000000]
ERROR(0): I-cache idx[0] tag[0000000000000000] utag[0000000000000000] stag[0000000000000000] u[0000000000000000] l[00\

ERROR(0): I-cache INSN0[0000000000000000] INSN1[0000000000000000] INSN2[0000000000000000] INSN3[0000000000000000]
ERROR(0): I-cache INSN4[0000000000000000] INSN5[0000000000000000] INSN6[0000000000000000] INSN7[0000000000000000]
ERROR(0): E-cache idx[800e40] tag[000000000e049f4c]
ERROR(0): E-cache data0[fffff8127d300180] data1[00000000004b5384] data2[0000000000000000] data3[0000000000000000]
Ker:xnel panic - not syncing: Irrecoverable deferred error trap.
----------------------------------------

Another quirk is that these copyarea calls will not happen until the
first drivers/char/vt.c:redraw_screen() occurs.  This will only happen
if you 1) VC switch or 2) run "consolechars" or 3) unblank the screen.

This seems to happen because until a redraw_screen() the screen scrolling
method used by fbcon is not finalized yet.  I've seen this with other fb
drivers too.

So if all you do is boot straight into X you will never see this bug on
the relevant chips.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agorelay: fix "full buffer with exactly full last subbuffer" accounting problem
Tom Zanussi [Wed, 6 Aug 2008 00:05:03 +0000 (00:05 +0000)]
relay: fix "full buffer with exactly full last subbuffer" accounting problem

commit 32194450330be327f3b25bf6b66298bd122599e9 upstream

In relay's current read implementation, if the buffer is completely full
but hasn't triggered the buffer-full condition (i.e. the last write
didn't cross the subbuffer boundary) and the last subbuffer is exactly
full, the subbuffer accounting code erroneously finds nothing available.
This patch fixes the problem.

Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Mathieu Desnoyers <compudj@krystal.dyndns.org>
Cc: Andrea Righi <righi.andrea@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoSCSI: block: Fix miscalculation of sg_io timeout in CDROM_SEND_PACKET handler.
Tim Wright [Tue, 5 Aug 2008 00:30:07 +0000 (00:30 +0000)]
SCSI: block: Fix miscalculation of sg_io timeout in CDROM_SEND_PACKET handler.

commit ad337591f4fd20de6a0ca03d6715267a5c1d2b16 upstream

It seems cdrwtool in the udftools has been unusable on "modern" kernels
for some time. A Google search reveals many people with the same issue
but no solution (cdrwtool fails to format the disk). After spending some
time tracking down the issue, it comes down to the following:

The udftools still use the older CDROM_SEND_PACKET interface to send
things like FORMAT_UNIT through to the drive. They should really be
updated, but that's another story. Since most distros are using libata
now, the cd or dvd burner appears as a SCSI device, and we wind up in
block/scsi_ioctl.c. Here, the code tries to take the "struct
cdrom_generic_command" and translate it and stuff it into a "struct
sg_io_hdr" structure so it can pass it to the modern sg_io() routine
instead. Unfortunately, there is one error, or rather an omission in the
translation. The timeout that is passed in in the "struct
cdrom_generic_command" is in HZ=100 units, and this is modified and
correctly converted to jiffies by use of clock_t_to_jiffies(). However,
a little further down, this cgc.timeout value in jiffies is simply
copied into the sg_io_hdr timeout, which should be in milliseconds.
Since most modern x86 kernels seems to be getting build with HZ=250, the
timeout that is passed to sg_io and eventually converted to the
timeout_per_command member of the scsi_cmnd structure is now four times
too small. Since cdrwtool tries to set the timeout to one hour for the
FORMAT_UNIT command, and it takes about 20 minutes to format a 4x CDRW,
the SCSI error-handler kicks in after the FORMAT_UNIT completes because
it took longer than the incorrectly-calculated timeout.

[jejb: fix up whitespace]
Signed-off-by: Tim Wright <timw@splhi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoSCSI: hptiop: add more PCI device IDs
HighPoint Linux Team [Tue, 5 Aug 2008 00:30:10 +0000 (00:30 +0000)]
SCSI: hptiop: add more PCI device IDs

commit dd07428b44944b42f699408fe31af47977f1e733 upstream

Add PCI device ID for new adapter models.

Signed-off-by: HighPoint Linux Team <linux@highpoint-tech.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoSCSI: scsi_transport_spi: fix oops in revalidate
James Bottomley [Tue, 5 Aug 2008 00:30:02 +0000 (00:30 +0000)]
SCSI: scsi_transport_spi: fix oops in revalidate

commit e8bac9e0647dd04c83fd0bfe7cdfe2f6dfb100d0 upstream

The class_device->device conversion is causing an oops in revalidate
because it's assuming that the device_for_each_child iterator will only
return struct scsi_device children.  The conversion made all former
class_devices children of the device as well, so this assumption is
broken.  Fix it.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoSCSI: ses: fix VPD inquiry overrun
James Bottomley [Tue, 5 Aug 2008 00:30:05 +0000 (00:30 +0000)]
SCSI: ses: fix VPD inquiry overrun

commit 671a99c8eb2f1dde08ac5538d8cd912047c61ddf upstream

There are a few kerneloops.org reports like this one:

http://www.kerneloops.org/search.php?search=ses_match_to_enclosure

That seem to imply we're running off the end of the VPD inquiry data
(although at 512 bytes, it should be long enough for just about
anything).  we should be using correctly sized buffers anyway, so put
those in and hope this oops goes away.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: ftdi_sio: add support for Luminance Stellaris Evaluation/Development Kits
Frederik Kriewitz [Thu, 14 Aug 2008 03:55:16 +0000 (03:55 +0000)]
USB: ftdi_sio: add support for Luminance Stellaris Evaluation/Development Kits

commit a00c3cadc2bf50b3c925acdb3d0e5789b1650498 upstream

The Patch adds support for Luminance Stellaris Evaluation/Development
Kits (FTDI 2232C based).
The PIDs were missing.

Successfully tested with a Stellaris LM3S8962 Evaluation kit.

Signed-off-by: Frederik Kriewitz <frederik@kriewitz.eu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: ftdi_sio: Add USB Product Id for ELV HS485
Andre Schenk [Thu, 14 Aug 2008 03:55:19 +0000 (03:55 +0000)]
USB: ftdi_sio: Add USB Product Id for ELV HS485

commit b5894a500127fce1db1309db5f9ca8b77a2ac266 upstream

USB product id registration for the ELV HS485 USB adapter (www.elv.de) to
their home automation bus system. Applies to 2.6.26.

Signed-off-by: Andre Schenk <andre@melior.s.bawue.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: pl2023: Remove USB id (4348:5523) handled by ch341
Tollef Fog Heen [Thu, 14 Aug 2008 03:55:05 +0000 (03:55 +0000)]
USB: pl2023: Remove USB id (4348:5523) handled by ch341

commit 8c809681ba0289afd0ed7bbb63679a0568dd441d upstream

USB ID 4348:5523 is handled by the ch341 driver.  Remove it from the
pl2023 driver.

Reverts 002e8f2c80c6be76bb312940bc278fc10b2b2487.

Signed-off-by: Tollef Fog Heen <tfheen@err.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agousb-serial: don't release unregistered minors
Alan Stern [Thu, 14 Aug 2008 03:55:11 +0000 (03:55 +0000)]
usb-serial: don't release unregistered minors

commit 0282b7f2a874e72c18fcd5a112ccf67f71ba7f5c upstream

This patch (as1121) fixes a bug in the USB serial core.  When a device
is unregistered, the core will give back its minors -- even if the
device hasn't been assigned any!

The patch reserves the highest minor value (255) to mean that no minor
was assigned.  It also removes some dead code and does a small style
fixup.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agousb-storage: unusual_devs entries for iRiver T10 and Datafab CF+SM reader
Alan Stern [Thu, 14 Aug 2008 03:55:07 +0000 (03:55 +0000)]
usb-storage: unusual_devs entries for iRiver T10 and Datafab CF+SM reader

commit 368ee6469c327364ea10082a348f91c1f5ba47f7 upstream

This patch (as1115) adds unusual_devs entries with the IGNORE_RESIDE
flag for the iRiver T10 and the Simple Tech/Datafab CF+SM card
reader.  Apparently these devices provide reasonable residue values
for READ and WRITE operations, but not for others like INQUIRY or READ
CAPACITY.

This fixes the iRiver T10 problem reported in Bugzilla #11125.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: usb-storage: quirk around v1.11 firmware on Nikon D4
Dave Jones [Thu, 14 Aug 2008 03:55:14 +0000 (03:55 +0000)]
USB: usb-storage: quirk around v1.11 firmware on Nikon D4

commit b9a097f26e55968cbc52e30a4a2e73d32d7604ce upstream

usb-storage: quirk around v1.11 firmware on Nikon D40

https://bugzilla.redhat.com/show_bug.cgi?id=454028
Just as in earlier firmware versions, we need to perform this
quirk for the latest version too.

Speculatively do the entry for the D80 too, as they seem to
have the same firmware problems historically.

Signed-off-by: Dave Jones <davej@redhat.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoLinux 2.6.25.15 v2.6.25.15
Greg Kroah-Hartman [Wed, 6 Aug 2008 17:11:47 +0000 (10:11 -0700)]
Linux 2.6.25.15

16 years agosound: ensure device number is valid in snd_seq_oss_synth_make_info
Willy Tarreau [Tue, 5 Aug 2008 00:20:03 +0000 (00:20 +0000)]
sound: ensure device number is valid in snd_seq_oss_synth_make_info

commit 82e68f7ffec3800425f2391c8c86277606860442 upstream

snd_seq_oss_synth_make_info() incorrectly reports information
to userspace without first checking for the validity of the
device number, leading to possible information leak (CVE-2008-3272).

Reported-By: Tobias Klein <tk@trapkit.de>
Acked-and-tested-by: Takashi Iwai <tiwai@suse.de>
Cc: stable@kernel.org
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agovfs: fix lookup on deleted directory
Miklos Szeredi [Wed, 2 Jul 2008 19:30:15 +0000 (21:30 +0200)]
vfs: fix lookup on deleted directory

commit d70b67c8bc72ee23b55381bd6a884f4796692f77 upstream

Lookup can install a child dentry for a deleted directory.  This keeps
the directory dentry alive, and the inode pinned in the cache and on
disk, even after all external references have gone away.

This isn't a big problem normally, since memory pressure or umount
will clear out the directory dentry and its children, releasing the
inode.  But for UBIFS this causes problems because its orphan area can
overflow.

Fix this by returning ENOENT for all lookups on a S_DEAD directory
before creating a child dentry.

Thanks to Zoltan Sogor for noticing this while testing UBIFS, and
Artem for the excellent analysis of the problem and testing.

Reported-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Tested-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: hda - Fix wrong volumes in AD1988 auto-probe mode
Takashi Iwai [Mon, 16 Jun 2008 13:47:26 +0000 (15:47 +0200)]
ALSA: hda - Fix wrong volumes in AD1988 auto-probe mode

commit 43785eaeb1cfb8aed3cf8027f298b242f88fdc45 upstream

Don't create mixer volume elements for Headphone and Speaker if they
use the same DAC as normal line-outs on AD1988.  Otherwise the amp
value gets screwed up, e.g.
https://bugzilla.novell.com/show_bug.cgi?id=398255

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: hda - Add missing Thinkpad Z60m support
Takashi Iwai [Mon, 30 Jun 2008 14:40:10 +0000 (16:40 +0200)]
ALSA: hda - Add missing Thinkpad Z60m support

commit 470eaf6be78424fc499a5039e5d5fe58bace2bc3 upstream

Added the missing SSID of Thinkpad Z60m for model=thinkpad with
AD1981HD.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: Fix Oops with usb-audio reconnection
Takashi Iwai [Thu, 17 Apr 2008 10:53:26 +0000 (12:53 +0200)]
ALSA: Fix Oops with usb-audio reconnection

Backport fixes for usb-audio Oops at reconnection.

9eb70e6... [ALSA] usb-audio - Fix race in reconnection
f18638d... [ALSA] Clean up snd_card_free*()
73d38b1... [ALSA] Fix the race of card instance unregistration

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: emu10k1 - Fix inverted Analog/Digital mixer switch on Audigy2
Takashi Iwai [Mon, 2 Jun 2008 09:45:53 +0000 (11:45 +0200)]
ALSA: emu10k1 - Fix inverted Analog/Digital mixer switch on Audigy2

commit d2cd74b158d7214a556226e3312f9fb1de64d7ae upstream

On Audigy2 Platinum, the Analog/Digital mixer switch is inverted.
https://bugzilla.novell.com/show_bug.cgi?id=396204

The patch adds a simple workaround.

There might be another device requiring a similar fix, too (or fix for
audigy2 generically), but right now I fix only the known broken one.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: Add more fallbacks to OSS PHONEOUT mixer map
Takashi Iwai [Mon, 4 Feb 2008 11:44:11 +0000 (12:44 +0100)]
ALSA: Add more fallbacks to OSS PHONEOUT mixer map

commit 6c4cc3a8ed15aacc06a5fd369639fef633cee2bc upstream

Added more fallbacks to OSS PHONEOUT mixer mapping.  This corresponds
to the speaker output in general, so now "Mono" and "Speaker" are
assigned.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: ac97 - Fix ASUS A9T laptop output
Takashi Iwai [Thu, 29 May 2008 06:16:56 +0000 (08:16 +0200)]
ALSA: ac97 - Fix ASUS A9T laptop output

commit e48d6d97bb6bd8c008045ea0522ea8278fdccc55 upstream

ASUS A9T laptop uses line-out pin as the real front-output while
other devices use it as the surround.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agopci: VT3336 can't do MSI either
Tejun Heo [Fri, 4 Jul 2008 16:59:32 +0000 (09:59 -0700)]
pci: VT3336 can't do MSI either

commit 66d715c95a39e84cd25204a665915621457d9691 upstream

It seems VT3336 can't do msi either as with its bro 3351.  Disable it.
Reported in the following SUSE bug.

  https://bugzilla.novell.com/show_bug.cgi?id=300001

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: EHCI: fix remote-wakeup regression
Alan Stern [Tue, 20 May 2008 20:58:58 +0000 (16:58 -0400)]
USB: EHCI: fix remote-wakeup regression

commit d1f114d12bb4db3147e1b1342ae31083c5a79c84 upstream

This patch (as1097) fixes a bug in the remote-wakeup handling in
ehci-hcd.  The driver currently does not keep track of whether the
change-suspend feature is enabled for each port; the feature is
automatically reset the first time it is read.  But recent changes to
the hub driver require that the feature be read at least twice in
order to work properly.

A bit-vector is added for storing the change-suspend feature values.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: appletouch - implement reset-resume logic
Oliver Neukum [Tue, 17 Jun 2008 15:56:55 +0000 (11:56 -0400)]
Input: appletouch - implement reset-resume logic

commit 90d95ef617a535a8832bdcb8dee07bf591e5dd82 upstream

On some boxes the touchpad needs to be reinitialized after resume to make
it function again. This fixes bugzilla #10825.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoACPI: update thermal temperature
Zhang, Rui [Thu, 10 Apr 2008 08:20:23 +0000 (16:20 +0800)]
ACPI: update thermal temperature

commit 76ecb4f2d7ea5c3aac8970b9529775316507c6d2 upstream

Fix the problem that thermal_get_temp returns the cached value,
which causes the temperature in generic thermal never updates.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoACPI: Reject below-freezing temperatures as invalid critical temperatures
Arjan van de Ven [Mon, 19 May 2008 22:55:15 +0000 (15:55 -0700)]
ACPI: Reject below-freezing temperatures as invalid critical temperatures

commit a39a2d7c72b358c6253a2ec28e17b023b7f6f41c upstream

My laptop thinks that it's a good idea to give -73C as the critical
CPU temperature.... which isn't the best thing since it causes a shutdown
right at bootup.

Temperatures below freezing are clearly invalid critical thresholds
so just reject these as such.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoAdd compat handler for PTRACE_GETSIGINFO
Andreas Schwab [Sat, 19 Apr 2008 16:25:13 +0000 (02:25 +1000)]
Add compat handler for PTRACE_GETSIGINFO

commit e4cc58944c1e2ce41e3079d4eb60c95e7ce04b2b upstream

Current versions of gdb require a working implementation of
PTRACE_GETSIGINFO for proper watchpoint support.  Since struct siginfo
contains pointers it must be converted when passed to a 32-bit debugger.

Signed-off-by: Andreas Schwab <schwab@suse.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoath5k: Use software encryption for now
Luis R. Rodriguez [Mon, 4 Feb 2008 02:53:20 +0000 (21:53 -0500)]
ath5k: Use software encryption for now

Commit 6844e63a9458d15b4437aa467c99128d994b0f6c

Hardware encryption doesn't work yet so lets use software
encryption for now.

Changes-licensed-under: 3-Clause-BSD

Signed-off-by: Luis R. Rodriguez <mcgrof@winlab.rutgers.edu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Cc: Jiri Benc <jbenc@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agobay: exit if notify handler cannot be installed
Holger Macht [Mon, 9 Jun 2008 23:22:24 +0000 (16:22 -0700)]
bay: exit if notify handler cannot be installed

commit 7efd52a407bed6a2b02015b8ebbff7beba155392 upstream

If acpi_install_notify_handler() for a bay device fails, the bay driver is
superfluous.  Most likely, another driver (like libata) is already caring
about this device anyway.  Furthermore,
register_hotplug_dock_device(acpi_handle) from the dock driver must not be
called twice with the same handler.  This would result in an endless loop
consuming 100% of CPU.  So clean up and exit.

Signed-off-by: Holger Macht <hmacht@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoBluetooth: Signal user-space for HIDP and BNEP socket errors
Marcel Holtmann [Mon, 14 Jul 2008 18:13:53 +0000 (20:13 +0200)]
Bluetooth: Signal user-space for HIDP and BNEP socket errors

commit ec8dab36e0738d3059980d144e34f16a26bbda7d upstream

When using the HIDP or BNEP kernel support, the user-space needs to
know if the connection has been terminated for some reasons. Wake up
the application if that happens. Otherwise kernel and user-space are
no longer on the same page and weird behaviors can happen.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agofat: detect media without partition table correctly
Frank Seidel [Mon, 28 Apr 2008 09:16:31 +0000 (02:16 -0700)]
fat: detect media without partition table correctly

commit 0607fd02587a6b4b086dc746d63123c1f284db68 upstream

I received a complaint that some FAT formated medias (e.g.  sd memory cards)
trigger a "unknown partition table" message even though there is no partition
table and they work correctly, while in general (when e.g.  formated with
mkdosfs or even Windows Vista) this message is not shown.

Currently this seems only to happen when the medias get formatted with Windows
XP (and possibly Win 2000).  Then the boot indicator byte contains garbage
(part of text message) and so do the other parts checked by msdos_paritition
which then later triggers this message.

References: novell bug #364365

Most fat formatted media without partition table contains zeros in the boot
indication and the other tested bytes and so falls through the checks in
msdos_partition, leading it to return with 1 (all is fine).

But some (e.g.  WinXP formatted) fat fomated medias don't use boot_ind and so
the check fails and causes a "unkown partition table" warning eventhough there
is none and everything would be fine.

This additional check directly verifies if there is a fat formatted medium
without a partition table.

Signed-off-by: Frank Seidel <fseidel@suse.de>
Cc: Andreas Dilger <adilger@sun.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoFAT_VALID_MEDIA(): remove pointless test
Andrew Morton [Mon, 28 Apr 2008 09:16:30 +0000 (02:16 -0700)]
FAT_VALID_MEDIA(): remove pointless test

commit 73f20e58b1d586e9f6d3ddc3aad872829aca7743 upstream

The on-disk media specification field in FAT is only 8-bits, so testing for
<=0xff is pointless, and can generate a "comparison is always true due to
limited range of data type" warning.

While we're there, convert FAT_VALID_MEDIA() into a C function - the present
implementation is buggy: it generates either one or two references to its
argument.

Cc: Frank Seidel <fseidel@suse.de>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: i8042 - add Acer Aspire 1360 to nomux blacklist
Jiri Kosina [Thu, 3 Jul 2008 14:45:38 +0000 (10:45 -0400)]
Input: i8042 - add Acer Aspire 1360 to nomux blacklist

commit 0376bce7b0659fe1e80d045860087072583ab93f upstream.

Acer Aspire 1360 needs to be added to nomux blacklist, otherwise its
touchpad misbehaves.

Reported-by: Clark Tompsett <clarkt@cnsp.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: i8042 - add Fujitsu-Siemens Amilo Pro 2010 to nomux list
Jiri Kosina [Wed, 30 Jan 2008 21:34:52 +0000 (16:34 -0500)]
Input: i8042 - add Fujitsu-Siemens Amilo Pro 2010 to nomux list

commit 3f79b1e94002791a42837a46b5817e87a0ac0873 upstream

Reported-by: Hans Aschauer <Hans.Aschauer@web.de>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: i8042 - add Fujitsu-Siemens Amilo Pro V2030 to nomux table
Jiri Kosina [Fri, 6 Jun 2008 04:56:43 +0000 (00:56 -0400)]
Input: i8042 - add Fujitsu-Siemens Amilo Pro V2030 to nomux table

Commit efd5184646d5d400fc538d093e9a0bec22a75551 upstream

Fujitsu Siemens Amilo Pro V2030 needs nomux table entry, in addition to
already existing entry for V2010 model (note that Fujitsu-Siemens changed
the capitalization in the DMI data for product).

Tested-by: Jiri Mleziva <jmleziva@tiscali.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: i8042 - add Gericom Bellagio to nomux blacklist
Jiri Kosina [Thu, 3 Jul 2008 15:00:28 +0000 (11:00 -0400)]
Input: i8042 - add Gericom Bellagio to nomux blacklist

commit 5b5b43d0b32ea586036638288c31179f00de5443 upstream

Gericom Bellagio needs to be added to nomux blacklist, otherwise its
touchpad misbehaves.

Reported-by: Roland Kletzing <roland.kletzing@materna.de>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: i8042 - add Intel D845PESV to nopnp list
Jiri Kosina [Wed, 28 May 2008 05:10:52 +0000 (01:10 -0400)]
Input: i8042 - add Intel D845PESV to nopnp list

commit c3a34f4390396a4bede3f8b7bcc5153f50b974bb upstream

This patch introduces i8042_dmi_nopnp_table to make it possible to perform
DMI matches for systems that need 'i8042.nopnp' to work correctly, and
introduces such an entry for Intel D845PESV -- this system doesn't
detect PS2 mouse reliably without this option, as reported by Robert
Lewis.

[dtor@mail.ru - make it compile if CONFIG_PNP is off - reported
 by Randy Dunlap]

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoInput: i8042 - retry failed CTR writes when resuming
Jiri Kosina [Tue, 17 Jun 2008 15:47:27 +0000 (11:47 -0400)]
Input: i8042 - retry failed CTR writes when resuming

commit 2f6a77d56523c14651236bc401a99b0e2aca2fdd upstream

There are systems that fail in i8042_resume() with

i8042: Can't write CTR to resume

as i8042_command(&i8042_ctr, I8042_CMD_CTL_WCTR) fails even though the
controller claimed itself to be ready before.

One retry after failing write fixes the problems on the failing systems.

Reported-by: Helmut Schaa <hschaa@novell.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agojbd: fix possible journal overflow issues
Josef Bacik [Mon, 28 Apr 2008 09:16:12 +0000 (02:16 -0700)]
jbd: fix possible journal overflow issues

commit 5b9a499d77e9dd39c9e6611ea10c56a31604f274 upstream

There are several cases where the running transaction can get buffers added to
its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers
counter end up larger than its t_outstanding_credits counter.

This will cause issues when starting new transactions as while we are logging
buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes
negative, we will report that we need less space in the journal than we
actually need, so transactions will be started even though there may not be
enough room for them.  In the worst case scenario (which admittedly is almost
impossible to reproduce) this will result in the journal running out of space.

The fix is to only
refile buffers from the committing transaction to the running transactions
BJ_Modified list when b_modified is set on that journal, which is the only way
to be sure if the running transaction has modified that buffer.

This patch also fixes an accounting error in journal_forget, it is possible
that we can call journal_forget on a buffer without having modified it, only
gotten write access to it, so instead of freeing a credit, we only do so if
the buffer was modified.  The assert will help catch if this problem occurs.
Without these two patches I could hit this assert within minutes of running
postmark, with them this issue no longer arises.  Thank you,

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Acked-by: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agojbd: fix race between free buffer and commit transaction
Mingming Cao [Fri, 25 Jul 2008 08:46:22 +0000 (01:46 -0700)]
jbd: fix race between free buffer and commit transaction

commit 3f31fddfa26b7594b44ff2b34f9a04ba409e0f91 upstream

journal_try_to_free_buffers() could race with jbd commit transaction when
the later is holding the buffer reference while waiting for the data
buffer to flush to disk.  If the caller of journal_try_to_free_buffers()
request tries hard to release the buffers, it will treat the failure as
error and return back to the caller.  We have seen the directo IO failed
due to this race.  Some of the caller of releasepage() also expecting the
buffer to be dropped when passed with GFP_KERNEL mask to the
releasepage()->journal_try_to_free_buffers().

With this patch, if the caller is passing the __GFP_WAIT and __GFP_FS to
indicating this call could wait, in case of try_to_free_buffers() failed,
let's waiting for journal_commit_transaction() to finish commit the
current committing transaction, then try to free those buffers again.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Badari Pulavarty <pbadari@us.ibm.com>
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agojbd: fix the way the b_modified flag is cleared
Josef Bacik [Mon, 28 Apr 2008 09:16:10 +0000 (02:16 -0700)]
jbd: fix the way the b_modified flag is cleared

commit 5bc833feaa8b2236265764e7e81f44937be46eda upstream

Currently at the start of a journal commit we loop through all of the buffers
on the committing transaction and clear the b_modified flag (the flag that is
set when a transaction modifies the buffer) under the j_list_lock.

The problem is that everywhere else this flag is modified only under the jbd
lock buffer flag, so it will race with a running transaction who could
potentially set it, and have it unset by the committing transaction.

This is also a big waste, you can have several thousands of buffers that you
are clearing the modified flag on when you may not need to.  This patch
removes this code and instead clears the b_modified flag upon entering
do_get_write_access/journal_get_create_access, so if that transaction does
indeed use the buffer then it will be accounted for properly, and if it does
not then we know we didn't use it.

That will be important for the next patch in this series.  Tested thoroughly
by myself using postmark/iozone/bonnie++.

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Acked-by: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoNFS: Ensure we zap only the access and acl caches when setting new acls
Trond Myklebust [Wed, 11 Jun 2008 21:39:04 +0000 (17:39 -0400)]
NFS: Ensure we zap only the access and acl caches when setting new acls

commit f41f741838480aeaa3a189cff6e210503cf9c42d upstream

...and ensure that we obey the NFS_INO_INVALID_ACL flag when retrieving the
acls.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoPOWERPC: PS3: Add time include to lpm
FUJITA Tomonori [Wed, 30 Apr 2008 22:25:09 +0000 (08:25 +1000)]
POWERPC: PS3: Add time include to lpm

commit 483d8876f75aa5707a646442377051f1b90db206 upstream

Add an include <asm/time.h> statement for get_tb().

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Geoff Levand <geoffrey.levand@am.sony.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Cc: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoreturn to old errno choice in mkdir() et.al.
Al Viro [Thu, 15 May 2008 08:49:12 +0000 (04:49 -0400)]
return to old errno choice in mkdir() et.al.

commit e9baf6e59842285bcf9570f5094e4c27674a0f7c upstream

In case when both EEXIST and EROFS would apply we used to
return the former in mkdir(2) and friends.  Lest anyone suspects
us of being consistent, in the same situation knfsd gave clients
nfs_erofs...

ro-bind series had switched the syscall side of things to
returning -EROFS and immediately broke an application - namely,
mkdir -p.  Patch restores the original behaviour...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Jan Blunck <jblunck@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoSCSI: megaraid_mbox: fix Dell CERC firmware problem
Hannes Reinecke [Fri, 18 Apr 2008 20:57:20 +0000 (13:57 -0700)]
SCSI: megaraid_mbox: fix Dell CERC firmware problem

commit 69cd39e94669e2994277a29249b6ef93b088ddbb upstream

Newer Dell CERC firmware (>= 6.62) implement a random deletion handling
compatible with the legacy megaraid driver.  The legacy handling shifted
the target ID by 0x80 only for I/O commands (READ/WRITE/etc), whereas
megaraid_mbox shifts the target ID always if random deletion is supported.
The resulted in megaraid_mbox sending an INQUIRY to the wrong channel, and
not finding any devices, obviously.

So we disable the random deletion support if the offending firmware is
found.

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=6695

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Bo Yang <Bo.Yang@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agox86: ioremap of 64-bit resource on 32-bit kernel fix
Ingo Molnar [Tue, 25 Mar 2008 06:31:17 +0000 (07:31 +0100)]
x86: ioremap of 64-bit resource on 32-bit kernel fix

commit 756a6c68556600aec9460346332884d891d5beb4 upstream

x86: ioremap of 64-bit resource on 32-bit kernel fix

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoromfs_readpage: don't report errors for pages beyond i_size
Linus Torvalds [Wed, 30 Jul 2008 21:26:25 +0000 (14:26 -0700)]
romfs_readpage: don't report errors for pages beyond i_size

commit 0056e65f9e28d83ee1a3fb4f7d0041e838f03c34 upstream

We zero-fill them like we are supposed to, and that's all fine.  It's
only an error if the 'romfs_copyfrom()' routine isn't able to fill the
data that is supposed to be there.

Most of the patch is really just re-organizing the code a bit, and using
separate variables for the error value and for how much of the page we
actually filled from the filesystem.

Reported-and-tested-by: Chris Fester <cfester@wms.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Matt Waddel <matt.waddel@freescale.com>
Cc: Greg Ungerer <gerg@snapgear.com>
Signed-of-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoLinux 2.6.25.14 v2.6.25.14
Greg Kroah-Hartman [Fri, 1 Aug 2008 19:03:00 +0000 (12:03 -0700)]
Linux 2.6.25.14

16 years agoFix off-by-one error in iov_iter_advance()
Linus Torvalds [Wed, 30 Jul 2008 22:20:18 +0000 (22:20 +0000)]
Fix off-by-one error in iov_iter_advance()

commit 94ad374a0751f40d25e22e036c37f7263569d24c upstream

The iov_iter_advance() function would look at the iov->iov_len entry
even though it might have iterated over the whole array, and iov was
pointing past the end.  This would cause DEBUG_PAGEALLOC to trigger a
kernel page fault if the allocation was at the end of a page, and the
next page was unallocated.

The quick fix is to just change the order of the tests: check that there
is any iovec data left before we check the iov entry itself.

Thanks to Alexey Dobriyan for finding this case, and testing the fix.

Reported-and-tested-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agonetfilter -stable: nf_conntrack_tcp: fix endless loop
Patrick McHardy [Thu, 17 Jul 2008 12:07:47 +0000 (14:07 +0200)]
netfilter -stable: nf_conntrack_tcp: fix endless loop

netfilter: nf_conntrack_tcp: fix endless loop

Upstream commit 6b69fe0:

When a conntrack entry is destroyed in process context and destruction
is interrupted by packet processing and the packet is an attempt to
reopen a closed connection, TCP conntrack tries to kill the old entry
itself and returns NF_REPEAT to pass the packet through the hook
again. This may lead to an endless loop: TCP conntrack repeatedly
finds the old entry, but can not kill it itself since destruction
is already in progress, but destruction in process context can not
complete since TCP conntrack is keeping the CPU busy.

Drop the packet in TCP conntrack if we can't kill the connection
ourselves to avoid this.

Reported by: hemao77@gmail.com [ Kernel bugzilla #11058 ]
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoCorrect hash flushing from huge_ptep_set_wrprotect()
David Gibson [Fri, 18 Jul 2008 05:55:49 +0000 (15:55 +1000)]
Correct hash flushing from huge_ptep_set_wrprotect()

Correct hash flushing from huge_ptep_set_wrprotect() [stable tree version]

A fix for incorrect flushing of the hash page table at fork() for
hugepages was recently committed as
86df86424939d316b1f6cfac1b6204f0c7dee317.  Without this fix, a process
can make a MAP_PRIVATE hugepage mapping, then fork() and have writes
to the mapping after the fork() pollute the child's version.

Unfortunately this bug also exists in the stable branch.  In fact in
that case copy_hugetlb_page_range() from mm/hugetlb.c calls
ptep_set_wrprotect() directly, the hugepage variant hook
huge_ptep_set_wrprotect() doesn't even exist.

The patch below is a port of the fix to the stable25/master branch.
It introduces a huge_ptep_set_wrprotect() call, but this is #defined
to be equal to ptep_set_wrprotect() unless the arch defines its own
version and sets __HAVE_ARCH_HUGE_PTEP_SET_WRPROTECT.

This arch preprocessor flag is kind of nasty, but it seems the sanest
way to introduce this fix with minimum risk of breaking other archs
for whom prep_set_wprotect() is suitable for hugepages.

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoath5k: don't enable MSI, we cannot handle it yet
Pavel Roskin [Wed, 30 Jul 2008 18:20:14 +0000 (18:20 +0000)]
ath5k: don't enable MSI, we cannot handle it yet

commit 256b152b005e319f985f50f2a910a75ba0def74f upstream

MSI is a nice thing, but we cannot enable it without changing the
interrupt handler.  If we do it, we break MSI capable hardware,
specifically AR5006 chipset.

Signed-off-by: Pavel Roskin <proski@gnu.org>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agob43legacy: Release mutex in error handling code
Julia Lawall [Wed, 30 Jul 2008 18:20:12 +0000 (18:20 +0000)]
b43legacy: Release mutex in error handling code

commit 4104863fb4a724723d1d5f3cba9d3c5084087e45 upstream

The mutex is released on a successful return, so it would seem that it
should be released on an error return as well.

The semantic patch finds this problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
expression l;
@@

mutex_lock(l);
.. when != mutex_unlock(l)
    when any
    when strict
(
if (...) { ... when != mutex_unlock(l)
+   mutex_unlock(l);
    return ...;
}
|
mutex_unlock(l);
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agocpufreq acpi: only call _PPC after cpufreq ACPI init funcs got called already
Thomas Renninger [Wed, 30 Jul 2008 18:20:10 +0000 (18:20 +0000)]
cpufreq acpi: only call _PPC after cpufreq ACPI init funcs got called already

commit a1531acd43310a7e4571d52e8846640667f4c74b upstream

Ingo Molnar provided a fix to not call _PPC at processor driver
initialization time in "[PATCH] ACPI: fix cpufreq regression" (git
commit e4233dec749a3519069d9390561b5636a75c7579)

But it can still happen that _PPC is called at processor driver
initialization time.

This patch should make sure that this is not possible anymore.

Signed-off-by: Thomas Renninger <trenn@suse.de>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Len Brown <lenb@kernel.org>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoeCryptfs: use page_alloc not kmalloc to get a page of memory
Eric Sandeen [Tue, 29 Jul 2008 02:50:12 +0000 (02:50 +0000)]
eCryptfs: use page_alloc not kmalloc to get a page of memory

commit 7fcba054373d5dfc43d26e243a5c9b92069972ee upstream
Date: Mon, 28 Jul 2008 15:46:39 -0700
Subject: eCryptfs: use page_alloc not kmalloc to get a page of memory

With SLUB debugging turned on in 2.6.26, I was getting memory corruption
when testing eCryptfs.  The root cause turned out to be that eCryptfs was
doing kmalloc(PAGE_CACHE_SIZE); virt_to_page() and treating that as a nice
page-aligned chunk of memory.  But at least with SLUB debugging on, this
is not always true, and the page we get from virt_to_page does not
necessarily match the PAGE_CACHE_SIZE worth of memory we got from kmalloc.

My simple testcase was 2 loops doing "rm -f fileX; cp /tmp/fileX ." for 2
different multi-megabyte files.  With this change I no longer see the
corruption.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Acked-by: Michael Halcrow <mhalcrow@us.ibm.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoixgbe: remove device ID for unsupported device
Jesse Brandeburg [Sat, 12 Jul 2008 00:34:58 +0000 (17:34 -0700)]
ixgbe: remove device ID for unsupported device

commit bb5d10ac8cc315d53306963001fe650d88a1cbb2 upstream

The ixgbe driver was untested with device ID 8086:10c8 but still advertises
support.  Currently if this device is present in the system when the driver
is loaded, the system will panic.
Remove this device ID until full support can be tested with available
hardware.  This patch is necessary for 2.6.24, 2.6.25 and 2.6.26

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agomarkers: fix markers read barrier for multiple probes
Mathieu Desnoyers [Wed, 30 Jul 2008 18:20:05 +0000 (18:20 +0000)]
markers: fix markers read barrier for multiple probes

commit 5def9a3a22e09c99717f41ab7f07ec9e1a1f3ec8 upstream

Paul pointed out two incorrect read barriers in the marker handler code in
the path where multiple probes are connected.  Those are ordering reads of
"ptype" (single or multi probe marker), "multi" array pointer, and "multi"
array data access.

It should be ordered like this :

read ptype
smp_rmb()
read multi array pointer
smp_read_barrier_depends()
access data referenced by multi array pointer

The code with a single probe connected (optimized case, does not have to
allocate an array) has correct memory ordering.

It applies to kernel 2.6.26.x, 2.6.25.x and linux-next.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agompc52xx_psc_spi: fix block transfer
Luotao Fu [Tue, 29 Jul 2008 02:50:14 +0000 (02:50 +0000)]
mpc52xx_psc_spi: fix block transfer

commit 9a7867e1b34c3575e7e76a05c0c54c6edbdae2a4 upstream

The block transfer routine in the mpc52xx psc spi driver misinterpret
the datasheet.  According to the processor datasheet the chipselect is
held as long as the EOF is not written.

Theoretically blocks of any sizes can be transferred in this way.  The
old routine however writes an EOF after every word, which has the size
of size_of_word.  This makes the transfer slow.

Also fixed some duplicate code.

Signed-off-by: Luotao Fu <l.fu@pengutronix.de>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agotmpfs: fix kernel BUG in shmem_delete_inode
Hugh Dickins [Tue, 29 Jul 2008 02:50:18 +0000 (02:50 +0000)]
tmpfs: fix kernel BUG in shmem_delete_inode

commit 14fcc23fdc78e9d32372553ccf21758a9bd56fa1 upstream

SuSE's insserve initscript ordering program hits kernel BUG at mm/shmem.c:814
on 2.6.26.  It's using posix_fadvise on directories, and the shmem_readpage
method added in 2.6.23 is letting POSIX_FADV_WILLNEED allocate useless pages
to a tmpfs directory, incrementing i_blocks count but never decrementing it.

Fix this by assigning shmem_aops (pointing to readpage and writepage and
set_page_dirty) only when it's needed, on a regular file or a long symlink.

Many thanks to Kel for outstanding bugreport and steps to reproduce it.

Reported-by: Kel Modderman <kel@otaku42.de>
Tested-by: Kel Modderman <kel@otaku42.de>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoVFS: increase pseudo-filesystem block size to PAGE_SIZE
Alex Nixon [Wed, 30 Jul 2008 18:20:08 +0000 (18:20 +0000)]
VFS: increase pseudo-filesystem block size to PAGE_SIZE

commit 3971e1a917548977cff71418a7c3575ffbc9571f upstream

This commit:

    commit ba52de123d454b57369f291348266d86f4b35070
    Author: Theodore Ts'o <tytso@mit.edu>
    Date:   Wed Sep 27 01:50:49 2006 -0700

        [PATCH] inode-diet: Eliminate i_blksize from the inode structure

caused the block size used by pseudo-filesystems to decrease from
PAGE_SIZE to 1024 leading to a doubling of the number of context switches
during a kernbench run.

Signed-off-by: Alex Nixon <Alex.Nixon@citrix.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ian Campbell <Ian.Campbell@eu.citrix.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agox86: fix kernel_physical_mapping_init() for large x86 systems
Ingo Molnar [Sun, 20 Jul 2008 15:24:03 +0000 (17:24 +0200)]
x86: fix kernel_physical_mapping_init() for large x86 systems

based on e22146e610bb7aed63282148740ab1d1b91e1d90 upstream

Fix bug in kernel_physical_mapping_init() that causes kernel
page table to be built incorrectly for systems with greater
than 512GB of memory.

Signed-off-by: Jack Steiner <steiner@sgi.com>
Cc: linux-mm@kvack.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Oliver Pinter <oliver.pntr@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoahci: retry enabling AHCI a few times before spitting out WARN_ON()
Tejun Heo [Fri, 18 Jul 2008 23:14:40 +0000 (19:14 -0400)]
ahci: retry enabling AHCI a few times before spitting out WARN_ON()

commit 15fe982e429e0e6b7466719acb6cfd9dbfe47f0c upstream

ahci: retry enabling AHCI a few times before spitting out WARN_ON()

Some chips need AHCI_EN set more than once to actually set it.  Try a
few times before giving up and spitting out WARN_ON().

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Peer Chen <pchen@nvidia.com>
Cc: Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Cc: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: trident - pause s/pdif output
Pierre Ossman [Fri, 18 Jul 2008 23:04:27 +0000 (19:04 -0400)]
ALSA: trident - pause s/pdif output

Commit 981bcead3f2279a1ec6fb5f2c57aff79ed61a700 upstream.

Stop the S/PDIF DMA engine and output when the device is told to pause.
It will keep on looping the current buffer contents if this isn't done.

Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Tested-by: Rene Herman <rene.herman@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Cc: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoALSA: hda - Fix "alc262_sony_unsol[]" hda_verb array
Akio Idehara [Sat, 26 Jul 2008 07:40:14 +0000 (16:40 +0900)]
ALSA: hda - Fix "alc262_sony_unsol[]" hda_verb array

[ALSA] hda - Fix "alc262_sony_unsol[]" hda_verb array

commit 7b1e8795ebfe1705153d1001f2a899119f4d9012 upstream

I think that hda_verb array must have "terminator (empty array)".
But alc262_sony_unsol[] does not have it.
And it causes gcc-4.3's buggy behavior
with snd_hda_sequence_write().

Signed-off-by: Akio Idehara <zbe64533@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoARM: fix fls() for 64-bit arguments
Andrew Morton [Sun, 27 Jul 2008 23:55:02 +0000 (23:55 +0000)]
ARM: fix fls() for 64-bit arguments

commit 0c65f459ce6c8bd873a61b3ae1e57858ab1debf3 upstream

arm's fls() is implemented as a macro, causing it to misbehave when passed
64-bit arguments.  Fix.

Cc: Nickolay Vinogradov <nickolay@protei.ru>
Tested-by: Krzysztof Halasa <khc@pm.waw.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoFix build on COMPAT platforms when CONFIG_EPOLL is disabled
Atsushi Nemoto [Tue, 22 Jul 2008 20:05:17 +0000 (20:05 +0000)]
Fix build on COMPAT platforms when CONFIG_EPOLL is disabled

commit 5f17156fc55abac476d180e480bedb0f07f01b14 upstream

Add missing cond_syscall() entry for compat_sys_epoll_pwait.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Cc: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoide-cd: fix oops when using growisofs
Jens Axboe [Thu, 24 Jul 2008 22:05:15 +0000 (22:05 +0000)]
ide-cd: fix oops when using growisofs

commit e8e7b9eb11c34ee18bde8b7011af41938d1ad667 upstream

cdrom_read_capacity() will blindly return the capacity from the device
without sanity-checking it.  This later causes code in fs/buffer.c to
oops.

Fix this by checking that the device is telling us sensible things.

From: Jens Axboe <jens.axboe@oracle.com>
Cc: Michael Buesch <mb@bu3sch.de>
Cc: Jan Kara <jack@suse.cz>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Borislav Petkov <petkovbb@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[bart: print device name instead of driver name]
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
[harvey: blocklen is a big-endian value]
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agomarkers: fix duplicate modpost entry
Mathieu Desnoyers [Tue, 22 Jul 2008 20:05:14 +0000 (20:05 +0000)]
markers: fix duplicate modpost entry

commit: d35cb360c29956510b2fe1a953bd4968536f7216

When a kernel was rebuilt, the previous Module.markers was not cleared.
It caused markers with different format strings to appear as duplicates
when a markers was changed.  This problem is present since
scripts/mod/modpost.c started to generate Module.markers, commit
b2e3e658b344c6bcfb8fb694100ab2f2b5b2edb0

It therefore applies to 2.6.25, 2.6.26 and linux-next.

I merely merged the patches from Roland, Wenji and Takashi here.

Credits to
Roland McGrath <roland@redhat.com>
Wenji Huang <wenji.huang@oracle.com>
and
Takashi Nishiie <t-nishiie@np.css.fujitsu.com>

for providing the individual fixes.

- Changelog :
  - Integrated Takashi's Makefile modification to clear Module.markers upon
    make clean.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Roland McGrath <roland@redhat.com>
Cc: Wenji Huang <wenji.huang@oracle.com>
Cc: Takashi Nishiie <t-nishiie@np.css.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agopata_atiixp: Don't disable
Alan Cox [Fri, 18 Jul 2008 23:13:02 +0000 (19:13 -0400)]
pata_atiixp: Don't disable

Commit 05177f178efe1459d2d0ac05430027ba201889a4 upstream

pata_atiixp: Don't disable

A couple of distributions (Fedora, Ubuntu) were having weird problems with the
ATI IXP series PATA controllers being reported as simplex.  At the heart of
the problem is that both distros ignored the recommendations to load pata_acpi
and ata_generic *AFTER* specific host drivers.

The underlying cause however is that if you D3 and then D0 an ATI IXP it
helpfully throws away some configuration and won't let you rewrite it.

Add checks to ata_generic and pata_acpi to pin ATIIXP devices.  Possibly the
real answer here is to quirk them and pin them, but right now we can't do that
before they've been pcim_enable()'d by a driver.

I'm indebted to David Gero for this.  His bug report not only reported the
problem but identified the cause correctly and he had tested the right values
to prove what was going on

[If you backport this for 2.6.24 you will need to pull in the 2.6.25
removal of the bogus WARN_ON() in pcim_enagle]

Signed-off-by: Alan Cox <alan@redhat.com>
Tested-by: David Gero <davidg@havidave.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agosparc64: Do not define BIO_VMERGE_BOUNDARY.
David S. Miller [Mon, 21 Jul 2008 20:17:38 +0000 (13:17 -0700)]
sparc64: Do not define BIO_VMERGE_BOUNDARY.

[ Upstream commit 74988bd85d1cb97987534fd7ffbc570e81145418 ]

The IOMMU code and the block layer can split things
up using different rules, so this can't work reliably.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agosparc64: Fix cpufreq notifier registry.
David S. Miller [Wed, 23 Jul 2008 23:21:07 +0000 (16:21 -0700)]
sparc64: Fix cpufreq notifier registry.

[ Upstream commit 7ae93f51d7fa8b9130d47e0b7d17979a165c5bc3 ]

Based upon a report by Daniel Smolik.

We do it too early, which triggers a BUG in
cpufreq_register_notifier().

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agosparc64: Fix lockdep issues in LDC protocol layer.
David S. Miller [Wed, 23 Jul 2008 05:34:29 +0000 (22:34 -0700)]
sparc64: Fix lockdep issues in LDC protocol layer.

[ Upstream commit b7c2a75725dee9b5643a0aae3a4cb47f52e00a49 ]

We're calling request_irq() with a IRQs disabled.

No straightforward fix exists because we want to
enable these IRQs and setup state atomically before
getting into the IRQ handler the first time.

What happens now is that we mark the VIRQ to not be
automatically enabled by request_irq().  Then we
make explicit enable_irq() calls when we grab the
LDC channel.

This way we don't need to call request_irq() illegally
under the LDC channel lock any more.

Bump LDC version and release date.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agotcp: Clear probes_out more aggressively in tcp_ack().
David S. Miller [Fri, 25 Jul 2008 23:53:38 +0000 (16:53 -0700)]
tcp: Clear probes_out more aggressively in tcp_ack().

[ Upstream commit 4b53fb67e385b856a991d402096379dab462170a ]

This is based upon an excellent bug report from Eric Dumazet.

tcp_ack() should clear ->icsk_probes_out even if there are packets
outstanding.  Otherwise if we get a sequence of ACKs while we do have
packets outstanding over and over again, we'll never clear the
probes_out value and eventually think the connection is too sick and
we'll reset it.

This appears to be some "optimization" added to tcp_ack() in the 2.4.x
timeframe.  In 2.2.x, probes_out is pretty much always cleared by
tcp_ack().

Here is Eric's original report:

----------------------------------------
Apparently, we can in some situations reset TCP connections in a couple of seconds when some frames are lost.

In order to reproduce the problem, please try the following program on linux-2.6.25.*

Setup some iptables rules to allow two frames per second sent on loopback interface to tcp destination port 12000

iptables -N SLOWLO
iptables -A SLOWLO -m hashlimit --hashlimit 2 --hashlimit-burst 1 --hashlimit-mode dstip --hashlimit-name slow2 -j ACCEPT
iptables -A SLOWLO -j DROP

iptables -A OUTPUT -o lo -p tcp --dport 12000 -j SLOWLO

Then run the attached program and see the output :

# ./loop
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,200ms,1)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,200ms,3)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,200ms,5)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,200ms,7)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,200ms,9)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,200ms,11)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,201ms,13)
State      Recv-Q Send-Q                                  Local Address:Port                                    Peer Address:Port
ESTAB      0      40                                          127.0.0.1:54455                                      127.0.0.1:12000  timer:(persist,188ms,15)
write(): Connection timed out
wrote 890 bytes but was interrupted after 9 seconds
ESTAB      0      0                 127.0.0.1:12000            127.0.0.1:54455
Exiting read() because no data available (4000 ms timeout).
read 860 bytes

While this tcp session makes progress (sending frames with 50 bytes of payload, every 500ms), linux tcp stack decides to reset it, when tcp_retries 2 is reached (default value : 15)

tcpdump :

15:30:28.856695 IP 127.0.0.1.56554 > 127.0.0.1.12000: S 33788768:33788768(0) win 32792 <mss 16396,nop,nop,sackOK,nop,wscale 7>
15:30:28.856711 IP 127.0.0.1.12000 > 127.0.0.1.56554: S 33899253:33899253(0) ack 33788769 win 32792 <mss 16396,nop,nop,sackOK,nop,wscale 7>
15:30:29.356947 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 1:61(60) ack 1 win 257
15:30:29.356966 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 61 win 257
15:30:29.866415 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 61:111(50) ack 1 win 257
15:30:29.866427 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 111 win 257
15:30:30.366516 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 111:161(50) ack 1 win 257
15:30:30.366527 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 161 win 257
15:30:30.876196 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 161:211(50) ack 1 win 257
15:30:30.876207 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 211 win 257
15:30:31.376282 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 211:261(50) ack 1 win 257
15:30:31.376290 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 261 win 257
15:30:31.885619 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 261:311(50) ack 1 win 257
15:30:31.885631 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 311 win 257
15:30:32.385705 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 311:361(50) ack 1 win 257
15:30:32.385715 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 361 win 257
15:30:32.895249 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 361:411(50) ack 1 win 257
15:30:32.895266 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 411 win 257
15:30:33.395341 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 411:461(50) ack 1 win 257
15:30:33.395351 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 461 win 257
15:30:33.918085 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 461:511(50) ack 1 win 257
15:30:33.918096 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 511 win 257
15:30:34.418163 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 511:561(50) ack 1 win 257
15:30:34.418172 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 561 win 257
15:30:34.927685 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 561:611(50) ack 1 win 257
15:30:34.927698 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 611 win 257
15:30:35.427757 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 611:661(50) ack 1 win 257
15:30:35.427766 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 661 win 257
15:30:35.937359 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 661:711(50) ack 1 win 257
15:30:35.937376 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 711 win 257
15:30:36.437451 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 711:761(50) ack 1 win 257
15:30:36.437464 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 761 win 257
15:30:36.947022 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 761:811(50) ack 1 win 257
15:30:36.947039 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 811 win 257
15:30:37.447135 IP 127.0.0.1.56554 > 127.0.0.1.12000: P 811:861(50) ack 1 win 257
15:30:37.447203 IP 127.0.0.1.12000 > 127.0.0.1.56554: . ack 861 win 257
15:30:41.448171 IP 127.0.0.1.12000 > 127.0.0.1.56554: F 1:1(0) ack 861 win 257
15:30:41.448189 IP 127.0.0.1.56554 > 127.0.0.1.12000: R 33789629:33789629(0) win 0

Source of program :

/*
 * small producer/consumer program.
 * setup a listener on 127.0.0.1:12000
 * Forks a child
 *   child connect to 127.0.0.1, and sends 10 bytes on this tcp socket every 100 ms
 * Father accepts connection, and read all data
 */
#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <unistd.h>
#include <stdio.h>
#include <time.h>
#include <sys/poll.h>

int port = 12000;
char buffer[4096];
int main(int argc, char *argv[])
{
        int lfd = socket(AF_INET, SOCK_STREAM, 0);
        struct sockaddr_in socket_address;
        time_t t0, t1;
        int on = 1, sfd, res;
        unsigned long total = 0;
        socklen_t alen = sizeof(socket_address);
        pid_t pid;

        time(&t0);
        socket_address.sin_family = AF_INET;
        socket_address.sin_port = htons(port);
        socket_address.sin_addr.s_addr = htonl(INADDR_LOOPBACK);

        if (lfd == -1) {
                perror("socket()");
                return 1;
        }
        setsockopt(lfd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(int));
        if (bind(lfd, (struct sockaddr *)&socket_address, sizeof(socket_address)) == -1) {
                perror("bind");
                close(lfd);
                return 1;
        }
        if (listen(lfd, 1) == -1) {
                perror("listen()");
                close(lfd);
                return 1;
        }
        pid = fork();
        if (pid == 0) {
                int i, cfd = socket(AF_INET, SOCK_STREAM, 0);
                close(lfd);
                if (connect(cfd, (struct sockaddr *)&socket_address, sizeof(socket_address)) == -1) {
                        perror("connect()");
                        return 1;
                        }
                for (i = 0 ; ;) {
                        res = write(cfd, "blablabla\n", 10);
                        if (res > 0) total += res;
                        else if (res == -1) {
                                perror("write()");
                                break;
                        } else break;
                        usleep(100000);
                        if (++i == 10) {
                                system("ss -on dst 127.0.0.1:12000");
                                i = 0;
                        }
                }
                time(&t1);
                fprintf(stderr, "wrote %lu bytes but was interrupted after %g seconds\n", total, difftime(t1, t0));
                system("ss -on | grep 127.0.0.1:12000");
                close(cfd);
                return 0;
        }
        sfd = accept(lfd, (struct sockaddr *)&socket_address, &alen);
        if (sfd == -1) {
                perror("accept");
                return 1;
        }
        close(lfd);
        while (1) {
                struct pollfd pfd[1];
                pfd[0].fd = sfd;
                pfd[0].events = POLLIN;
                if (poll(pfd, 1, 4000) == 0) {
                        fprintf(stderr, "Exiting read() because no data available (4000 ms timeout).\n");
                        break;
                }
                res = read(sfd, buffer, sizeof(buffer));
                if (res > 0) total += res;
                else if (res == 0) break;
                else perror("read()");
        }
        fprintf(stderr, "read %lu bytes\n", total);
        close(sfd);
        return 0;
}
----------------------------------------

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agovmlinux.lds: move __attribute__((__cold__)) functions back into final .text section
Jan Beulich [Sun, 27 Jul 2008 17:30:04 +0000 (17:30 +0000)]
vmlinux.lds: move __attribute__((__cold__)) functions back into final .text section

commit fb5e2b379732e1a6ea32392980bb42e0212db842 upstream

Due to the addition of __attribute__((__cold__)) to a few symbols
without adjusting the linker scripts, those symbols currently may end
up outside the [_stext,_etext) range, as they get placed in
.text.unlikely by (at least) gcc 4.3.0. This may confuse code not only
outside of the kernel, symbol_put_addr()'s BUG() could also trigger.
Hence we need to add .text.unlikely (and for future uses of
__attribute__((__hot__)) also .text.hot) to the TEXT_TEXT() macro.

Issue observed by Lukas Lipavsky.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Tested-by: Lukas Lipavsky <llipavsky@suse.cz>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agox86: fix crash due to missing debugctlmsr on AMD K6-3
Jan Kratochvil [Tue, 22 Jul 2008 20:55:04 +0000 (20:55 +0000)]
x86: fix crash due to missing debugctlmsr on AMD K6-3

commit d536b1f86591fb081c7a56eab04e711eb4dab951 upstream

currently if you use PTRACE_SINGLEBLOCK on AMD K6-3 (i586) it will crash.
Kernel now wrongly assumes existing DEBUGCTLMSR MSR register there.

Removed the assumption also for some other non-K6 CPUs but I am not sure there
(but it can only bring small inefficiency there if my assumption is wrong).

Based on info from Roland McGrath, Chuck Ebbert and Mikulas Patocka.
More info at:
https://bugzilla.redhat.com/show_bug.cgi?id=456175

Signed-off-by: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoisofs: fix minor filesystem corruption
Adam Greenblatt [Fri, 25 Jul 2008 18:15:11 +0000 (18:15 +0000)]
isofs: fix minor filesystem corruption

commit c0a1633b6201ef79e31b7da464d44fdf5953054d upstream

Some iso9660 images contain files with rockridge data that is either
incorrect or incompletely parsed.  Prior to commit
f2966632a134e865db3c819346a1dc7d96e05309 ("[PATCH] rock: handle directory
overflows") (included with kernel 2.6.13) the kernel ignored the rockridge
data for these files, while still allowing the files to be accessed under
their non-rockridge names.  That commit inadvertently changed things so
that files with invalid rockridge data could not be accessed at all.  (I
ran across the problem when comparing some old CDs with hard disk copies I
had made long ago under kernel 2.4: a few of the files on the hard disk
copies were no longer visible on the CDs.)

This change reverts to the pre-2.6.13 behavior.

Signed-off-by: Adam Greenblatt <adam.greenblatt@gmail.com>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoquota: fix possible infinite loop in quota code
Jan Kara [Fri, 25 Jul 2008 18:15:07 +0000 (18:15 +0000)]
quota: fix possible infinite loop in quota code

commit b48d380541f634663b71766005838edbb7261685 upstream

When quota structure is going to be dropped and it is dirty, quota code tries
to write it.  If the write fails for some reason (e.  g.  transaction cannot
be started because the journal is aborted), we try writing again and again and
again...  Fix the problem by clearing the dirty bit even if the write failed.

(akpm: for 2.6.27, 2.6.26.x and 2.6.25.x)

Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: dingdinghua <dingdinghua85@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoLinux 2.6.25.13 v2.6.25.13
Greg Kroah-Hartman [Mon, 28 Jul 2008 17:59:18 +0000 (10:59 -0700)]
Linux 2.6.25.13

16 years agoudplite: Protection against coverage value wrap-around
Gerrit Renker [Thu, 24 Jul 2008 07:11:56 +0000 (00:11 -0700)]
udplite: Protection against coverage value wrap-around

[ Upstream commit 47112e25da41d9059626033986dc3353e101f815 ]

This patch clamps the cscov setsockopt values to a maximum of 0xFFFF.

Setsockopt values greater than 0xffff can cause an unwanted
wrap-around.  Further, IPv6 jumbograms are not supported (RFC 3838,
3.5), so that values greater than 0xffff are not even useful.

Further changes: fixed a typo in the documentation.

[ Add USHORT_MAX from upstream to linux/kernel.h -DaveM ]

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoxfrm: fix fragmentation for ipv4 xfrm tunnel
Steffen Klassert [Thu, 24 Jul 2008 06:55:40 +0000 (23:55 -0700)]
xfrm: fix fragmentation for ipv4 xfrm tunnel

[ Upstream commit fe833fca2eac6b3d3ad5e35f44ad4638362f1da8 ]

When generating the ip header for the transformed packet we just copy
the frag_off field of the ip header from the original packet to the ip
header of the new generated packet. If we receive a packet as a chain
of fragments, all but the last of the new generated packets have the
IP_MF flag set. We have to mask the frag_off field to only keep the
IP_DF flag from the original packet. This got lost with git commit
36cf9acf93e8561d9faec24849e57688a81eb9c5 ("[IPSEC]: Separate
inner/outer mode processing on output")

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoraw: Restore /proc/net/raw correct behavior
Eric Dumazet [Thu, 24 Jul 2008 06:54:35 +0000 (23:54 -0700)]
raw: Restore /proc/net/raw correct behavior

[ Upstream commit 68be802cd5ad040fe8cfa33ce3031405df2d9117 ]

I just noticed "cat /proc/net/raw" was buggy, missing '\n' separators.

I believe this was introduced by commit 8cd850efa4948d57a2ed836911cfd1ab299e89c6
([RAW]: Cleanup IPv4 raw_seq_show.)

This trivial patch restores correct behavior, and applies to current
Linus tree (should also be applied to stable tree as well.)

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>