Add bf60x cpu pm callbacks and change blackfin pm framework to support bf60x.
Signed-off-by: NSteven Miao <realmz6@gmail.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

bf60x support l2 hardware ecc error check, add panic when double bits ecc error
happened.
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Signed-off-by: NSonic Zhang <sonic.zhang@analog.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Signed-off-by: NSteven Miao <realmz6@gmail.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Signed-off-by: NScott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Bf60x support big CPLB pages, this commit enable it.
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Add system event controller support for bf60x so that interrupt can be
handled.
Signed-off-by: NSteven Miao <realmz6@gmail.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Add clock support for bf60x.
Signed-off-by: NSteven Miao <realmz6@gmail.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

This patch added bf60x to current blackfin kernel framework.
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Add machine files for bf60x including head files, Kconfig/Makefile and board
file.
Signed-off-by: NSonic Zhang <sonic.zhang@analog.com>
Signed-off-by: NBob Liu <lliubbo@gmail.com>

do{...}while makes no sense in __ASSEMBLY__ code paths.  now
kernels fail to build:
arch/blackfin/mach-bf561/atomic.S: Assembler messages:
arch/blackfin/mach-bf561/atomic.S:48: Error: syntax error. Input text
was do.
arch/blackfin/mach-bf561/atomic.S:48: Error:
arch/blackfin/mach-bf561/atomic.S:48: Error: syntax error. Input text
was }.
arch/blackfin/mach-bf561/atomic.S:48: Error:
arch/blackfin/mach-bf561/atomic.S:53: Error: syntax error. Input text
was do.
arch/blackfin/mach-bf561/atomic.S:53: Error:
arch/blackfin/mach-bf561/atomic.S:53: Error: syntax error. Input text
was }.
Signed-off-by: NBob Liu <lliubbo@gmail.com>

SMP support was removed from BF561-EZKIT-SMP_defconfig in last commit by
mistake, this patch add it back.
Signed-off-by: NBob Liu <lliubbo@gmail.com>

Pull PA-RISC fixes from James Bottomley:
 "This is a set of three bug fixes that gets parisc running again on
  systems with PA1.1 processors.

  Two fix regressions introduced in 2.6.39 and one fixes a prefetch bug
  that only affects PA7300LC processors.  We also have another pending
  fix to do with the sectional arrangement of vmlinux.lds, but there's a
  query on it during testing on one particular system type, so I'll hold
  off sending it in for now."

* tag 'parisc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/parisc-2.6:
  [PARISC] fix panic on prefetch(NULL) on PA7300LC
  [PARISC] fix crash in flush_icache_page_asm on PA1.1
  [PARISC] fix PA1.1 oops on boot

Pull x86 linker bug workarounds from Peter Anvin.

GNU ld-2.22.52.0.[12] (*) has an unfortunate bug where it incorrectly
turns certain relocation entries absolute.  Section-relative symbols
that are part of otherwise empty sections are silently changed them to
absolute.  We rely on section-relative symbols staying section-relative,
and actually have several sections in the linker script solely for this
purpose.

See for example

   http://sourceware.org/bugzilla/show_bug.cgi?id=14052

We could just black-list the buggy linker, but it appears that it got
shipped in at least F17, and possibly other distros too, so it's sadly
not some rare unusual case.

This backports the workaround from the x86/trampoline branch, and as
Peter says: "This is not a minimal fix, not at all, but it is a tested
code base."

* 'x86/ld-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86, relocs: When printing an error, say relative or absolute
  x86, relocs: Workaround for binutils 2.22.52.0.1 section bug
  x86, realmode: 16-bit real-mode code support for relocs tool

(*) That's a manly release numbering system. Stupid, sure. But manly.

Pull block layer fixes from Jens Axboe:
 "A few small, but important fixes.  Most of them are marked for stable
  as well

   - Fix failure to release a semaphore on error path in mtip32xx.
   - Fix crashable condition in bio_get_nr_vecs().
   - Don't mark end-of-disk buffers as mapped, limit it to i_size.
   - Fix for build problem with CONFIG_BLOCK=n on arm at least.
   - Fix for a buffer overlow on UUID partition printing.
   - Trivial removal of unused variables in dac960."

* 'for-linus' of git://git.kernel.dk/linux-block:
  block: fix buffer overflow when printing partition UUIDs
  Fix blkdev.h build errors when BLOCK=n
  bio allocation failure due to bio_get_nr_vecs()
  block: don't mark buffers beyond end of disk as mapped
  mtip32xx: release the semaphore on an error path
  dac960: Remove unused variables from DAC960_CreateProcEntries()

Pull one more networking bug-fix from David Miller:
 "One last straggler.

  Eric Dumazet's pktgen unload oops fix was not entirely complete, but
  all the cases should be handled properly now....  fingers crossed."

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  pktgen: fix module unload for good

Occasionally, testing memcg's move_charge_at_immigrate on rc7 shows
a flurry of hundreds of warnings at kernel/res_counter.c:96, where
res_counter_uncharge_locked() does WARN_ON(counter->usage < val).

The first trace of each flurry implicates __mem_cgroup_cancel_charge()
of mc.precharge, and an audit of mc.precharge handling points to
mem_cgroup_move_charge_pte_range()'s THP handling in commit 12724850
("memcg: avoid THP split in task migration").

Checking !mc.precharge is good everywhere else, when a single page is to
be charged; but here the "mc.precharge -= HPAGE_PMD_NR" likely to
follow, is liable to result in underflow (a lot can change since the
precharge was estimated).

Simply check against HPAGE_PMD_NR: there's probably a better
alternative, trying precharge for more, splitting if unsuccessful; but
this one-liner is safer for now - no kernel/res_counter.c:96 warnings
seen in 26 hours.
Signed-off-by: NHugh Dickins <hughd@google.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When the relocs tool throws an error, let the error message say if it
is an absolute or relative symbol.  This should make it a lot more
clear what action the programmer needs to take and should help us find
the reason if additional symbol bugs show up.
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>
Cc: <stable@vger.kernel.org>

GNU ld 2.22.52.0.1 has a bug that it blindly changes symbols from
section-relative to absolute if they are in a section of zero length.
This turns the symbols __init_begin and __init_end into absolute
symbols.  Let the relocs program know that those should be treated as
relative symbols.
Reported-by: NIngo Molnar <mingo@kernel.org>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>
Cc: H.J. Lu <hjl.tools@gmail.com>
Cc: <stable@vger.kernel.org>
Cc: Jarkko Sakkinen <jarkko.sakkinen@intel.com>

A new option is added to the relocs tool called '--realmode'.
This option causes the generation of 16-bit segment relocations
and 32-bit linear relocations for the real-mode code. When
the real-mode code is moved to the low-memory during kernel
initialization, these relocation entries can be used to
relocate the code properly.

In the assembly code 16-bit segment relocations must be relative
to the 'real_mode_seg' absolute symbol. Linear relocations must be
relative to a symbol prefixed with 'pa_'.

16-bit segment relocation is used to load cs:ip in 16-bit code.
Linear relocations are used in the 32-bit code for relocatable
data references. They are declared in the linker script of the
real-mode code.

The relocs tool is moved to arch/x86/tools/relocs.c, and added new
target archscripts that can be used to build scripts needed building
an architecture.  be compiled before building the arch/x86 tree.

[ hpa: accelerating this because it detects invalid absolute
  relocations, a serious bug in binutils 2.22.52.0.x which currently
  produces bad kernels. ]
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Link: http://lkml.kernel.org/r/1336501366-28617-2-git-send-email-jarkko.sakkinen@intel.comSigned-off-by: NJarkko Sakkinen <jarkko.sakkinen@intel.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: <stable@vger.kernel.org>

Pull a dm fix from Alasdair G Kergon:
 "A fix to the thin provisioning userspace interface."

* tag 'dm-3.4-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm:
  dm thin: fix table output when pool target disables discard passdown internally

When the thin pool target clears the discard_passdown parameter
internally, it incorrectly changes the table line reported to userspace.
This breaks dumb string comparisons on these table lines in generic
userspace device-mapper library code and leads to tables being reloaded
repeatedly when nothing is actually meant to be changing.

This patch corrects this by no longer changing the table line when
discard passdown was disabled.

We can still tell when discard passdown is overridden by looking for the
message "Discard unsupported by data device (sdX): Disabling discard passdown."

This automatic detection is also moved from the 'load' to the 'resume'
so that it is re-evaluated should the properties of underlying devices
change.
Signed-off-by: NMike Snitzer <snitzer@redhat.com>
Acked-by: NJoe Thornber <ejt@redhat.com>
Signed-off-by: NAlasdair G Kergon <agk@redhat.com>

Pull one more md bugfix from NeilBrown:
 "Fix bug in recent fix to RAID10.

  Without this patch, recovery will crash"

* tag 'md-3.4-fixes' of git://neil.brown.name/md:
  md/raid10: fix transcription error in calc_sectors conversion.

Pull tile tree bugfix from Chris Metcalf:
 "This fixes a security vulnerability (and correctness bug) in tilegx"

* 'stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  tilegx: enable SYSCALL_WRAPPERS support

The old code was
		sector_div(stride, fc);
the new code was
		sector_dir(size, conf->near_copies);

'size' is right (the stride various wasn't really needed), but
'fc' means 'far_copies', and that is an important difference.

Signed-off-by: NeilBrown <neilb@suse.de>       

Merge misc fixes from Andrew Morton.

* emailed from Andrew Morton <akpm@linux-foundation.org>: (4 patches)
  frv: delete incorrect task prototypes causing compile fail
  slub: missing test for partial pages flush work in flush_all()
  fs, proc: fix ABBA deadlock in case of execution attempt of map_files/ entries
  drivers/rtc/rtc-pl031.c: configure correct wday for 2000-01-01

Instead of doing the i_mode calculations at proc_fd_instantiate() time,
move them into tid_fd_revalidate(), which is where the other inode state
(notably uid/gid information) is updated too.

Otherwise we'll end up with stale i_mode information if an fd is re-used
while the dentry still hangs around.  Not that anything really *cares*
(symlink permissions don't really matter), but Tetsuo Handa noticed that
the owner read/write bits don't always match the state of the
readability of the file descriptor, and we _used_ to get this right a
long time ago in a galaxy far, far away.

Besides, aside from fixing an ugly detail (that has apparently been this
way since commit 61a28784: "proc: Remove the hard coded inode
numbers" in 2006), this removes more lines of code than it adds.  And it
just makes sense to update i_mode in the same place we update i_uid/gid.

Al Viro correctly points out that we could just do the inode fill in the
inode iops ->getattr() function instead.  However, that does require
somewhat slightly more invasive changes, and adds yet *another* lookup
of the file descriptor.  We need to do the revalidate() for other
reasons anyway, and have the file descriptor handy, so we might as well
fill in the information at this point.
Reported-by: NTetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: NEric Biederman <ebiederm@xmission.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

commit c57b5468 (pktgen: fix crash at module unload) did a very poor
job with list primitives.

1) list_splice() arguments were in the wrong order

2) list_splice(list, head) has undefined behavior if head is not
initialized.

3) We should use the list_splice_init() variant to clear pktgen_threads
list.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Some discussion with the glibc mailing lists revealed that this was
necessary for 64-bit platforms with MIPS-like sign-extension rules
for 32-bit values.  The original symptom was that passing (uid_t)-1 to
setreuid() was failing in programs linked -pthread because of the "setxid"
mechanism for passing setxid-type function arguments to the syscall code.
SYSCALL_WRAPPERS handles ensuring that all syscall arguments end up with
proper sign-extension and is thus the appropriate fix for this problem.

On other platforms (s390, powerpc, sparc64, and mips) this was fixed
in 2.6.28.6.  The general issue is tracked as CVE-2009-0029.

Cc: <stable@vger.kernel.org>
Signed-off-by: NChris Metcalf <cmetcalf@tilera.com>

Pull a machine check recovery fix from Tony Luck.

I really don't like how the MCE code does some of the things it does,
but this does seem to be an improvement.

* tag 'linus-mce-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
  x86/mce: Only restart instruction after machine check recovery if it is safe

Commit 41101809 ("fork: Provide weak arch_release_[task_struct|
thread_info] functions") in -tip highlights a problem in the frv arch,
where it has needles prototypes for alloc_task_struct_node and
free_task_struct.  This now shows up as:

  kernel/fork.c:120:66: error: static declaration of 'alloc_task_struct_node' follows non-static declaration
  kernel/fork.c:127:51: error: static declaration of 'free_task_struct' follows non-static declaration

since that commit turned them into real functions.  Since arch/frv does
does not define define __HAVE_ARCH_TASK_STRUCT_ALLOCATOR (i.e.  it just
uses the generic ones) it shouldn't list these at all.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

I found some kernel messages such as:

    SLUB raid5-md127: kmem_cache_destroy called for cache that still has objects.
    Pid: 6143, comm: mdadm Tainted: G           O 3.4.0-rc6+        #75
    Call Trace:
    kmem_cache_destroy+0x328/0x400
    free_conf+0x2d/0xf0 [raid456]
    stop+0x41/0x60 [raid456]
    md_stop+0x1a/0x60 [md_mod]
    do_md_stop+0x74/0x470 [md_mod]
    md_ioctl+0xff/0x11f0 [md_mod]
    blkdev_ioctl+0xd8/0x7a0
    block_ioctl+0x3b/0x40
    do_vfs_ioctl+0x96/0x560
    sys_ioctl+0x91/0xa0
    system_call_fastpath+0x16/0x1b

Then using kmemleak I found these messages:

    unreferenced object 0xffff8800b6db7380 (size 112):
      comm "mdadm", pid 5783, jiffies 4294810749 (age 90.589s)
      hex dump (first 32 bytes):
        01 01 db b6 ad 4e ad de ff ff ff ff ff ff ff ff  .....N..........
        ff ff ff ff ff ff ff ff 98 40 4a 82 ff ff ff ff  .........@J.....
      backtrace:
        kmemleak_alloc+0x21/0x50
        kmem_cache_alloc+0xeb/0x1b0
        kmem_cache_open+0x2f1/0x430
        kmem_cache_create+0x158/0x320
        setup_conf+0x649/0x770 [raid456]
        run+0x68b/0x840 [raid456]
        md_run+0x529/0x940 [md_mod]
        do_md_run+0x18/0xc0 [md_mod]
        md_ioctl+0xba8/0x11f0 [md_mod]
        blkdev_ioctl+0xd8/0x7a0
        block_ioctl+0x3b/0x40
        do_vfs_ioctl+0x96/0x560
        sys_ioctl+0x91/0xa0
        system_call_fastpath+0x16/0x1b

This bug was introduced by commit a8364d55 ("slub: only IPI CPUs that
have per cpu obj to flush"), which did not include checks for per cpu
partial pages being present on a cpu.
Signed-off-by: Nmajianpeng <majianpeng@gmail.com>
Cc: Gilad Ben-Yossef <gilad@benyossef.com>
Acked-by: NChristoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Tested-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

map_files/ entries are never supposed to be executed, still curious
minds might try to run them, which leads to the following deadlock

  ======================================================
  [ INFO: possible circular locking dependency detected ]
  3.4.0-rc4-24406-g841e6a6 #121 Not tainted
  -------------------------------------------------------
  bash/1556 is trying to acquire lock:
   (&sb->s_type->i_mutex_key#8){+.+.+.}, at: do_lookup+0x267/0x2b1

  but task is already holding lock:
   (&sig->cred_guard_mutex){+.+.+.}, at: prepare_bprm_creds+0x2d/0x69

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #1 (&sig->cred_guard_mutex){+.+.+.}:
         validate_chain+0x444/0x4f4
         __lock_acquire+0x387/0x3f8
         lock_acquire+0x12b/0x158
         __mutex_lock_common+0x56/0x3a9
         mutex_lock_killable_nested+0x40/0x45
         lock_trace+0x24/0x59
         proc_map_files_lookup+0x5a/0x165
         __lookup_hash+0x52/0x73
         do_lookup+0x276/0x2b1
         walk_component+0x3d/0x114
         do_last+0xfc/0x540
         path_openat+0xd3/0x306
         do_filp_open+0x3d/0x89
         do_sys_open+0x74/0x106
         sys_open+0x21/0x23
         tracesys+0xdd/0xe2

  -> #0 (&sb->s_type->i_mutex_key#8){+.+.+.}:
         check_prev_add+0x6a/0x1ef
         validate_chain+0x444/0x4f4
         __lock_acquire+0x387/0x3f8
         lock_acquire+0x12b/0x158
         __mutex_lock_common+0x56/0x3a9
         mutex_lock_nested+0x40/0x45
         do_lookup+0x267/0x2b1
         walk_component+0x3d/0x114
         link_path_walk+0x1f9/0x48f
         path_openat+0xb6/0x306
         do_filp_open+0x3d/0x89
         open_exec+0x25/0xa0
         do_execve_common+0xea/0x2f9
         do_execve+0x43/0x45
         sys_execve+0x43/0x5a
         stub_execve+0x6c/0xc0

This is because prepare_bprm_creds grabs task->signal->cred_guard_mutex
and when do_lookup happens we try to grab task->signal->cred_guard_mutex
again in lock_trace.

Fix it using plain ptrace_may_access() helper in proc_map_files_lookup()
and in proc_map_files_readdir() instead of lock_trace(), the caller must
be CAP_SYS_ADMIN granted anyway.
Signed-off-by: NCyrill Gorcunov <gorcunov@openvz.org>
Reported-by: NSasha Levin <levinsasha928@gmail.com>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Dave Jones <davej@redhat.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The reset date of the ST Micro version of PL031 is 2000-01-01.  The
correct weekday for 2000-01-01 is saturday, but pl031 is initialized to
sunday.  This may lead to alarm malfunction, so configure the correct
wday if RTC_DR indicates reset.
Signed-off-by: NRajkumar Kasirajan <rajkumar.kasirajan@stericsson.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>
Cc: Mattias Wallin <mattias.wallin@stericsson.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Pull ARM fixes from Russell King:
 "Small set of fixes again."

* 'fixes' of git://git.linaro.org/people/rmk/linux-arm:
  ARM: 7419/1: vfp: fix VFP flushing regression on sigreturn path
  ARM: 7418/1: LPAE: fix access flag setup in mem_type_table
  ARM: prevent VM_GROWSDOWN mmaps extending below FIRST_USER_ADDRESS
  ARM: 7417/1: vfp: ensure preemption is disabled when enabling VFP access

Pull two networking fixes from David S. Miller:

1) Thanks to Willy Tarreau and Eric Dumazet, we've unlocked a bug that's
   been present in do_tcp_sendpages() since that function was written in
   2002.

   When we block to wait for memory we have to unconditionally try and
   push out pending TCP data, otherwise we can block for an unreasonably
   long amount of time.

2) Fix deadlock in e1000, fixes kernel bugzilla 43132

   From Tushar Dave.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  e1000: Prevent reset task killing itself.
  tcp: do_tcp_sendpages() must try to push data out on oom conditions

Commit 1cc0c998 ("ACPI: Fix D3hot v D3cold confusion") introduced a
bug in __acpi_bus_set_power() and changed the behavior of
acpi_pci_set_power_state() in such a way that it generally doesn't work
as expected if PCI_D3hot is passed to it as the second argument.

First off, if ACPI_STATE_D3 (equal to ACPI_STATE_D3_COLD) is passed to
__acpi_bus_set_power() and the explicit_set flag is set for the D3cold
state, the function will try to execute AML method called "_PS4", which
doesn't exist.

Fix this by adding a check to ensure that the name of the AML method
to execute for transitions to ACPI_STATE_D3_COLD is correct in
__acpi_bus_set_power().  Also make sure that the explicit_set flag
for ACPI_STATE_D3_COLD will be set if _PS3 is present and modify
acpi_power_transition() to avoid accessing power resources for
ACPI_STATE_D3_COLD, because they don't exist.

Second, if PCI_D3hot is passed to acpi_pci_set_power_state() as the
target state, the function will request a transition to
ACPI_STATE_D3_HOT instead of ACPI_STATE_D3.  However,
ACPI_STATE_D3_HOT is now only marked as supported if the _PR3 AML
method is defined for the given device, which is rare.  This causes
problems to happen on systems where devices were successfully put
into ACPI D3 by pci_set_power_state(PCI_D3hot) which doesn't work
now.  In particular, some unused graphics adapters are not turned
off as a result.

To fix this issue restore the old behavior of
acpi_pci_set_power_state(), which is to request a transition to
ACPI_STATE_D3 (equal to ACPI_STATE_D3_COLD) if either PCI_D3hot or
PCI_D3cold is passed to it as the argument.

This approach is not ideal, because generally power should not
be removed from devices if PCI_D3hot is the target power state,
but since this behavior is relied on, we have no choice but to
restore it at the moment and spend more time on designing a
better solution in the future.

References: https://bugzilla.kernel.org/show_bug.cgi?id=43228Reported-by: Nrocko <rockorequin@hotmail.com>
Reported-by: NCristian Rodríguez <crrodriguez@opensuse.org>
Reported-and-tested-by: NPeter <lekensteyn@gmail.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Killing reset task while adapter is resetting causes deadlock.
Only kill reset task if adapter is not resetting.
Ref bug #43132 on bugzilla.kernel.org

CC: stable@vger.kernel.org
Signed-off-by: NTushar Dave <tushar.n.dave@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since recent changes on TCP splicing (starting with commits 2f533844
"tcp: allow splice() to build full TSO packets" and 35f9c09f "tcp:
tcp_sendpages() should call tcp_push() once"), I started seeing
massive stalls when forwarding traffic between two sockets using
splice() when pipe buffers were larger than socket buffers.

Latest changes (net: netdev_alloc_skb() use build_skb()) made the
problem even more apparent.

The reason seems to be that if do_tcp_sendpages() fails on out of memory
condition without being able to send at least one byte, tcp_push() is not
called and the buffers cannot be flushed.

After applying the attached patch, I cannot reproduce the stalls at all
and the data rate it perfectly stable and steady under any condition
which previously caused the problem to be permanent.

The issue seems to have been there since before the kernel migrated to
git, which makes me think that the stalls I occasionally experienced
with tux during stress-tests years ago were probably related to the
same issue.

This issue was first encountered on 3.0.31 and 3.2.17, so please backport
to -stable.
Signed-off-by: NWilly Tarreau <w@1wt.eu>
Acked-by: NEric Dumazet <edumazet@google.com>
Cc: <stable@vger.kernel.org>