Paul Mundt [Fri, 11 Jan 2008 04:18:16 +0000 (13:18 +0900)]
sh: Force __access_ok() to obey address space limit.
When the thread_info->addr_limit changes were introduced, __access_ok()
was missed in the conversion, allowing user processes to perform P1/P2
accesses under certain conditions.
This has already been corrected with the nommu refactoring in later
kernels.
Carmelo Amoroso [Mon, 7 Jan 2008 04:50:18 +0000 (13:50 +0900)]
sh: Fix argument page dcache flushing regression.
In the do_execve() path, argument page handling used to explicitly call
flush_dcache_page() for each page, this has since been reworked and
uses flush_kernel_dcache_page() instead, which is presently a nop.
Doing a simple modprobe/rmmod in a loop under busybox consistently
manages to crash without providing a sane flush_kernel_dcache_page()
implementation, so, plug in a simple implementation.
Signed-off-by: Carmelo Amoroso <carmelo73@gmail.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Ingo Molnar [Sun, 30 Dec 2007 10:58:17 +0000 (11:58 +0100)]
CPU hotplug: fix cpu_is_offline() on !CONFIG_HOTPLUG_CPU
make randconfig bootup testing found that the cpufreq code
crashes on bootup, if the powernow-k8 driver is enabled and
if maxcpus=1 passed on the boot line to a !CONFIG_HOTPLUG_CPU
kernel.
First lockdep found out that there's an inconsistent unlock
sequence:
=====================================
[ BUG: bad unlock balance detected! ]
-------------------------------------
swapper/1 is trying to release lock (&per_cpu(cpu_policy_rwsem, cpu)) at:
[<ffffffff806ffd8e>] unlock_policy_rwsem_write+0x3c/0x42
but there are no more locks to release!
then shortly afterwards the cpufreq code crashed on an assert:
------------[ cut here ]------------
kernel BUG at drivers/cpufreq/cpufreq.c:1068!
invalid opcode: 0000 [1] SMP
[...]
Call Trace:
[<ffffffff805145d6>] sysdev_driver_unregister+0x5b/0x91
[<ffffffff806ff520>] cpufreq_register_driver+0x15d/0x1a2
[<ffffffff80cc0596>] powernowk8_init+0x86/0x94
[...]
---[ end trace 1e9219be2b4431de ]---
the bug was caused by maxcpus=1 bootup, which brought up the
secondary core as !cpu_online() but !cpu_is_offline() either,
which on on !CONFIG_HOTPLUG_CPU is always 0 (include/linux/cpu.h):
/* CPUs don't go offline once they're online w/o CONFIG_HOTPLUG_CPU */
static inline int cpu_is_offline(int cpu) { return 0; }
but the cpufreq code uses cpu_online() and cpu_is_offline() in
a mixed way - the low-level drivers use cpu_online(), while
the cpufreq core uses cpu_is_offline(). This opened up the
possibility to add the non-initialized sysdev device of the
secondary core:
cpufreq-core: trying to register driver powernow-k8
cpufreq-core: adding CPU 0
powernow-k8: BIOS error - no PSB or ACPI _PSS objects
cpufreq-core: initialization failed
cpufreq-core: adding CPU 1
cpufreq-core: initialization failed
which then blew up. The fix is to make cpu_is_offline() always
the negation of cpu_online(). With that fix applied the kernel
boots up fine without crashing:
Calling initcall 0xffffffff80cc0510: powernowk8_init+0x0/0x94()
powernow-k8: Found 1 AMD Athlon(tm) 64 X2 Dual Core Processor 3800+ processors (1 cpu cores) (version 2.20.00)
powernow-k8: BIOS error - no PSB or ACPI _PSS objects
initcall 0xffffffff80cc0510: powernowk8_init+0x0/0x94() returned -19.
initcall 0xffffffff80cc0510 ran for 19 msecs: powernowk8_init+0x0/0x94()
Calling initcall 0xffffffff80cc328f: init_lapic_nmi_sysfs+0x0/0x39()
We could fix this by making CPU enumeration aware of max_cpus, but that
would be more fragile IMO, and the cpu_online(cpu) != cpu_is_offline(cpu)
possibility was quite confusing and a continuous source of bugs too.
Most distributions have kernels with CPU hotplug enabled, so this bug
remained hidden for a long time.
Bug forensics:
The broken cpu_is_offline() API variant was introduced via:
Ingo Molnar [Sun, 6 Jan 2008 20:09:53 +0000 (21:09 +0100)]
hda_intel suspend latency: shorten codec read
not sleeping for every codec read/write but doing a short udelay and
a conditional reschedule has cut suspend+resume latency by about 1
second on my T60.
The patch also fixes the unexpected codec-connection errors that
happen more often in the new power-save mode:
http://lkml.org/lkml/2007/11/8/255
http://bugzilla.kernel.org/show_bug.cgi?id=9332
Akos Maroy [Sun, 6 Jan 2008 10:15:55 +0000 (11:15 +0100)]
fix: using joysticks in 32 bit applications on 64 bit systems
unfortunately 32 bit apps don't see the joysticks on a 64 bit system.
this prevents one playing X-Plane (http://www.x-plane.com/) or other
32-bit games with joysticks.
this is a known issue, and already raised several times:
It turns out that the bug that was triggered by that commit was
apparently not actually triggered by that commit at all, and just the
testing conditions had changed enough to make it appear to be due to it.
The real problem seems to have been found by Peter Osterlund:
"pktcdvd sets it [block device size] when opening the /dev/pktcdvd
device, but when the drive is later opened as /dev/scd0, there is
nothing that sets it back. (Btw, 40944 is possible if the disk is a
CDRW that was formatted with "cdrwtool -m 10236".)
The problem is that pktcdvd opens the cd device in non-blocking mode
when pktsetup is run, and doesn't close it again until pktsetup -d is
run. The effect is that if you meanwhile open the cd device,
blkdev.c:do_open() doesn't call bd_set_size() because
bdev->bd_openers is non-zero."
" 1. Start with an empty drive.
2. pktsetup 0 /dev/scd0
3. Insert a CD containing an isofs filesystem.
4. mount /dev/pktcdvd/0 /mnt/tmp
5. umount /mnt/tmp
6. Press the eject button.
7. Insert a DVD containing a non-writable filesystem.
8. mount /dev/scd0 /mnt/tmp
9. find /mnt/tmp -type f -print0 | xargs -0 sha1sum >/dev/null
10. If the DVD contains data beyond the physical size of a CD, you
get I/O errors in the terminal, and dmesg reports lots of
"attempt to access beyond end of device" errors."
which in turn is because the nested open after the media change won't
cause the size to be set properly (because the original open still holds
the block device, and we only do the bd_set_size() when we don't have
other people holding the device open).
The proper fix for that is probably to just do something like
in fs/block_dev.c:do_open() even for the cases where we're not the
original opener (but *not* call bd_set_size(), since that will also
change the block size of the device).
Cc: Peter Osterlund <petero2@telia.com> Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Cc: Matthew Wilcox <matthew@wil.cx> Cc: Ingo Molnar <mingo@elte.hu> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dave Dillow [Fri, 4 Jan 2008 02:34:49 +0000 (21:34 -0500)]
[SCSI] SRP transport: only remove our own entries
The SCSI SRP transport class currently iterates over all children
devices of the host that is being removed in srp_remove_host(). However,
not all of those children were created by the SRP transport, and
removing them will cause corruption and an oops when their creator tries
to remove them.
Signed-off-by: David Dillow <dillowda@ornl.gov> Acked-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
[ISDN]: i4l: Fix DLE handling for i4l-audio
[ISDN] i4l: 'NO CARRIER' message lost after ldisc flush
[CONNECTOR]: Return proper error code in cn_call_callback()
[INET]: Fix netdev renaming and inet address labels
[CASSINI]: Bump driver version and release date.
[CASSINI]: Fix two obvious NAPI bugs.
[CASSINI]: Set skb->truesize properly on receive packets.
[CASSINI]: Program parent Intel31154 bridge when necessary.
[CASSINI]: Revert 'dont touch page_count'.
[CASSINI]: Fix endianness bug.
[XFRM]: Do not define km_migrate() if !CONFIG_XFRM_MIGRATE
[X25]: Add missing x25_neigh_put
Matthias Goebl [Fri, 4 Jan 2008 11:45:28 +0000 (03:45 -0800)]
[ISDN]: i4l: Fix DLE handling for i4l-audio
The DLE handling in i4l-audio seems to be broken.
It produces spurious DLEs so asterisk 1.2.24 with chan_modem_i4l
gets irritated, the error message is:
"chan_modem_i4l.c:450 i4l_read: Value of escape is ^ (17)".
-> There shouldn't be a DLE-^.
If a spurious DLE-ETX occurs, the audio connection even dies.
I use a "AVM Fritz!PCI" isdn card.
I found two issues that only appear if ISDN_AUDIO_SKB_DLECOUNT(skb) > 0:
- The loop in isdn_tty.c:isdn_tty_try_read() doesn't escape a DLE if it's
the last character.
- The loop in isdn_common.c:isdn_readbchan_tty() doesn't copy its characters,
it only remembers the last one ("last = *p;").
Compare it with the loop in isdn_common.c:isdn_readbchan(), that *does*
copy them ("*cp++ = *p;") correctly.
The special handling of the "last" character made it more difficult.
I compared it to linux-2.4.19: There was no "last"-handling and both loops
did escape and copy all characters correctly.
Signed-off-by: Matthias Goebl <matthias.goebl@goebl.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Matthias Goebl [Fri, 4 Jan 2008 10:19:30 +0000 (02:19 -0800)]
[ISDN] i4l: 'NO CARRIER' message lost after ldisc flush
The ISDN tty layer doesn't produce a 'NO CARRIER' message after hangup.
I suppose it broke when tty_buffer_flush() has been added to
tty_ldisc_flush() in the commit below.
For isdn_tty_modem_result(RESULT_NO_CARRIER..) the
message inserted via isdn_tty_at_cout() -> tty_insert_flip_char()
is flushed immediately by tty_ldisc_flush() -> tty_buffer_flush().
More annoyingly, the audio abort sequence DLE-ETX is also lost.
This patch fixes only active audio connections, because I assume that nobody
changes the line discipline for audio.
For non-audio connections the problem remains.
Maybe we can remove the tty_ldisc_flush() in isdn_tty_modem_result()
at all because it's done at tty_close?
On Mon, May 07, 2007 at 04:05:57PM -0500, Paul Fulghum wrote:
> Flush the tty flip buffer when the line discipline
> input queue is flushed, including the user call
> tcflush(TCIFLUSH/TCIOFLUSH). This prevents unexpected
> stale data after a user application calls tcflush().
>
> Cc: Alan Cox <alan@lxorguk.org.uk>
> Cc: Antonino Ingargiola <tritemio@gmail.com>
> Signed-off-by: Paul Fulghum <paulkf@microgate.com>
>
> --- a/drivers/char/tty_io.c 2007-05-04 05:46:55.000000000 -0500
> +++ b/drivers/char/tty_io.c 2007-05-05 03:23:46.000000000 -0500
> @@ -1240,6 +1263,7 @@ void tty_ldisc_flush(struct tty_struct *
> ld->flush_buffer(tty);
> tty_ldisc_deref(ld);
> }
> + tty_buffer_flush(tty);
[..]
Signed-off-by: Matthias Goebl <matthias.goebl@goebl.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Mark McLoughlin [Fri, 4 Jan 2008 08:56:25 +0000 (00:56 -0800)]
[INET]: Fix netdev renaming and inet address labels
When re-naming an interface, the previous secondary address
labels get lost e.g.
$> brctl addbr foo
$> ip addr add 192.168.0.1 dev foo
$> ip addr add 192.168.0.2 dev foo label foo:00
$> ip addr show dev foo | grep inet
inet 192.168.0.1/32 scope global foo
inet 192.168.0.2/32 scope global foo:00
$> ip link set foo name bar
$> ip addr show dev bar | grep inet
inet 192.168.0.1/32 scope global bar
inet 192.168.0.2/32 scope global bar:2
Turns out to be a simple thinko in inetdev_changename() - clearly we
want to look at the address label, rather than the device name, for
a suffix to retain.
Signed-off-by: Mark McLoughlin <markmc@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Al Viro [Fri, 4 Jan 2008 02:49:00 +0000 (18:49 -0800)]
[CASSINI]: Fix endianness bug.
Here's proposed fix for RX checksum handling in cassini; it affects
little-endian working with half-duplex gigabit, but obviously needs
testing on big-endian too.
The problem is, we need to convert checksum to fixed-endian *before*
correcting for (unstripped) FCS. On big-endian it won't matter
(conversion is no-op), on little-endian it will, but only if FCS is
not stripped by hardware; i.e. in half-duplex gigabit mode when
->crc_size is set.
cassini.c part is that fix, cassini.h one consists of trivial
endianness annotations. With that applied the sucker is endian-clean,
according to sparse.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 4 Jan 2008 04:43:21 +0000 (20:43 -0800)]
[XFRM]: Do not define km_migrate() if !CONFIG_XFRM_MIGRATE
In include/net/xfrm.h we find :
#ifdef CONFIG_XFRM_MIGRATE
extern int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
struct xfrm_migrate *m, int num_bundles);
...
#endif
We can also guard the function body itself in net/xfrm/xfrm_state.c
with same condition.
(Problem spoted by sparse checker)
make C=2 net/xfrm/xfrm_state.o
...
net/xfrm/xfrm_state.c:1765:5: warning: symbol 'km_migrate' was not declared. Should it be static?
...
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Julia Lawall [Wed, 2 Jan 2008 03:30:30 +0000 (19:30 -0800)]
[X25]: Add missing x25_neigh_put
The function x25_get_neigh increments a reference count. At the point of
the second goto out, the result of calling x25_get_neigh is only stored in
a local variable, and thus no one outside the function will be able to
decrease the reference count. Thus, x25_neigh_put should be called before
the return in this case.
The problem was found using the following semantic match.
(http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@@
type T,T1,T2;
identifier E;
statement S;
expression x1,x2,x3;
int ret;
@@
T E;
...
* if ((E = x25_get_neigh(...)) == NULL)
S
... when != x25_neigh_put(...,(T1)E,...)
when != if (E != NULL) { ... x25_neigh_put(...,(T1)E,...); ...}
when != x1 = (T1)E
when != E = x3;
when any
if (...) {
... when != x25_neigh_put(...,(T2)E,...)
when != if (E != NULL) { ... x25_neigh_put(...,(T2)E,...); ...}
when != x2 = (T2)E
(
* return;
|
* return ret;
)
}
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk> Signed-off-by: David S. Miller <davem@davemloft.net>
Russ Anderson [Thu, 3 Jan 2008 16:23:49 +0000 (10:23 -0600)]
[IA64] Update Altix nofault code
Montecito and Montvale behaves slightly differently than previous
Itanium processors, resulting in the MCA due to a failed PIO read
to sometimes surfacing outside the nofault code. This code is
based on discussions with Intel CPU architects and verified at
customer sites.
Signed-off-by: Russ Anderson <rja@sgi.com> Signed-off-by: Tony Luck <tony.luck@intel.com>
Samuel Thibault [Thu, 3 Jan 2008 16:17:54 +0000 (16:17 +0000)]
Console is utf-8 by default
The console is now by default in UTF-8 mode. Fix the documentation on
the default value, so that we can explain behaviour that otherwise
causes bug-reports like this:
http://bugzilla.kernel.org/show_bug.cgi?id=9319
Also add the needed "vt." prefix, so that the boot-time config options
to switch back to the legacy 8-bit mode is actually documented
correctly.
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Thu, 3 Jan 2008 19:59:59 +0000 (11:59 -0800)]
Merge git://git.linux-nfs.org/pub/linux/nfs-2.6
* git://git.linux-nfs.org/pub/linux/nfs-2.6:
NFSv4: Fix open_to_lock_owner sequenceid allocation...
NFSv4: nfs4_open_confirm must not set the open_owner as confirmed on error
NFS: add newline to kernel warning message in auth_gss code
NFSv4: Fix circular locking dependency in nfs4_kill_renewd
NFS: Fix a possible Oops in fs/nfs/super.c
NFSv4 file locking is currently completely broken since it doesn't respect
the OPEN sequencing when it is given an unconfirmed lock_owner and needs to
do an open_to_lock_owner. Worse: it breaks the sunrpc rules by doing a
GFP_KERNEL allocation inside an rpciod callback.
Fix is to preallocate the open seqid structure in nfs4_alloc_lockdata if we
see that the lock_owner is unconfirmed.
Then, in nfs4_lock_prepare() we wait for either the open_seqid, if
the lock_owner is still unconfirmed, or else fall back to waiting on the
standard lock_seqid.
Trond Myklebust [Wed, 2 Jan 2008 20:19:18 +0000 (15:19 -0500)]
NFSv4: nfs4_open_confirm must not set the open_owner as confirmed on error
RFC3530 states that the open_owner is confirmed if and only if the client
sends an OPEN_CONFIRM request with the appropriate sequence id and stateid
within the lease period.
Looking at the code, it would seem that taking the clp->cl_sem in
nfs4_kill_renewd is completely redundant, since we're already guaranteed to
have exclusive access to the nfs_client (we're shutting down).
Trond Myklebust [Wed, 2 Jan 2008 18:28:57 +0000 (13:28 -0500)]
NFS: Fix a possible Oops in fs/nfs/super.c
Sigh... commit 4584f520e1f773082ef44ff4f8969a5d992b16ec (NFS: Fix NFS
mountpoint crossing...) had a slight flaw: server can be NULL if sget()
returned an existing superblock.
Fix the fix by dereferencing s->s_fs_info.
Thanks to Coverity/Adrian Bunk and Frank Filz for spotting the bug.
(See http://bugzilla.kernel.org/show_bug.cgi?id=9647)
Also add in the same namespace Oops fix for NFSv4 in both the mountpoint
crossing case, and the referral case.
Jean Delvare [Thu, 3 Jan 2008 12:33:31 +0000 (07:33 -0500)]
hwmon: (w83627ehf) Be more careful when changing VID input level
The VID input level change has been reported to cause trouble. Be more
careful in this respect:
* Only change the level on the W83627EHF/EHG. The W83627DHG is more
complex in this respect.
* Don't change the level if the VID pins are in output mode.
* Only set the level to TTL if VRM 9.x is used.
Signed-off-by: Jean Delvare <khali@linux-fr.org> Signed-off-by: Mark M. Hoffman <mhoffman@lightlink.com>
Al Viro [Wed, 2 Jan 2008 14:09:57 +0000 (14:09 +0000)]
restrict reading from /proc/<pid>/maps to those who share ->mm or can ptrace pid
Contents of /proc/*/maps is sensitive and may become sensitive after
open() (e.g. if target originally shares our ->mm and later does exec
on suid-root binary).
Check at read() (actually, ->start() of iterator) time that mm_struct
we'd grabbed and locked is
- still the ->mm of target
- equal to reader's ->mm or the target is ptracable by reader.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Acked-by: Rik van Riel <riel@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ingo Molnar [Wed, 2 Jan 2008 16:25:34 +0000 (17:25 +0100)]
scsi: revert "[SCSI] Get rid of scsi_cmnd->done"
This reverts commit 6f5391c283d7fdcf24bf40786ea79061919d1e1d ("[SCSI]
Get rid of scsi_cmnd->done") that was supposed to be a cleanup commit,
but apparently it causes regressions:
Bug 9370 - v2.6.24-rc2-409-g9418d5d: attempt to access beyond end of device
http://bugzilla.kernel.org/show_bug.cgi?id=9370
this patch should be reintroduced in a more split-up form to make
testing of it easier.
Signed-off-by: Ingo Molnar <mingo@elte.hu> Acked-by: Matthew Wilcox <matthew@wil.cx> Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Wed, 2 Jan 2008 21:04:48 +0000 (13:04 -0800)]
Unify /proc/slabinfo configuration
Both SLUB and SLAB really did almost exactly the same thing for
/proc/slabinfo setup, using duplicate code and per-allocator #ifdef's.
This just creates a common CONFIG_SLABINFO that is enabled by both SLUB
and SLAB, and shares all the setup code. Maybe SLOB will want this some
day too.
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
James Bottomley [Wed, 2 Jan 2008 17:14:30 +0000 (11:14 -0600)]
[SCSI] scsi_sysfs: restore prep_fn when ULD is removed
A recent bug report:
http://bugzilla.kernel.org/show_bug.cgi?id=9674
Was caused because the ULDs now set their own prep functions, but
don't necessarily reset the prep function back to the SCSI default
when they are removed. This leads to panics if commands are sent to
the device after the module is removed because the prep_fn is still
pointing to the old module code. The fix for this is to implement a
bus remove method that resets the prep_fn pointer correctly before
calling the ULD specific driver remove method.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Paul Mackerras [Wed, 2 Jan 2008 04:56:30 +0000 (15:56 +1100)]
[POWERPC] Fix build failure on Cell when CONFIG_SPU_FS=y
Commit aed3a8c9bb1a8623a618232087c5ff62718e3b9a introduced a
definition of notify_spus_active in .../cell/spu_syscalls.c, and
another definition under #ifndef MODULE in .../cell/spufs/sched.c.
The latter is not necessary and causes the build to fail when
CONFIG_SPU_FS=y, so this removes it. It also removes the export
of do_notify_spus_active, which is unnecessary.
Signed-off-by: Paul Mackerras <paulus@samba.org> Acked-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Jeremy Kerr <jk@ozlabs.org>
On the second use after module load, the kernel crashes.
This fixes the immediate problem (accessed and dirty bits not set as
expected in pmd_none_or_clear_bad). I can't see why this would cause
a crash, but I haven't been able to reproduce it once this is applied.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ingo Molnar [Sun, 30 Dec 2007 16:24:35 +0000 (17:24 +0100)]
sched: fix gcc warnings
Meelis Roos reported these warnings on sparc64:
CC kernel/sched.o
In file included from kernel/sched.c:879:
kernel/sched_debug.c: In function 'nsec_high':
kernel/sched_debug.c:38: warning: comparison of distinct pointer types lacks a cast
the debug check in do_div() is over-eager here, because the long long
is always positive in these places. Mark this by casting them to
unsigned long long.
no change in code output:
text data bss dec hex filename
51471 6582 376 58429 e43d sched.o.before
51471 6582 376 58429 e43d sched.o.after
Gavin McCullagh [Sun, 30 Dec 2007 03:11:21 +0000 (19:11 -0800)]
[TCP]: use non-delayed ACK for congestion control RTT
When a delayed ACK representing two packets arrives, there are two RTT
samples available, one for each packet. The first (in order of seq
number) will be artificially long due to the delay waiting for the
second packet, the second will trigger the ACK and so will not itself
be delayed.
According to rfc1323, the SRTT used for RTO calculation should use the
first rtt, so receivers echo the timestamp from the first packet in
the delayed ack. For congestion control however, it seems measuring
delayed ack delay is not desirable as it varies independently of
congestion.
The patch below causes seq_rtt and last_ackt to be updated with any
available later packet rtts which should have less (and hopefully
zero) delack delay. The rtt value then gets passed to
ca_ops->pkts_acked().
Where TCP_CONG_RTT_STAMP was set, effort was made to supress RTTs from
within a TSO chunk (!fully_acked), using only the final ACK (which
includes any TSO delay) to generate RTTs. This patch removes these
checks so RTTs are passed for each ACK to ca_ops->pkts_acked().
For non-delay based congestion control (cubic, h-tcp), rtt is
sometimes used for rtt-scaling. In shortening the RTT, this may make
them a little less aggressive. Delay-based schemes (eg vegas, veno,
illinois) should get a cleaner, more accurate congestion signal,
particularly for small cwnds. The congestion control module can
potentially also filter out bad RTTs due to the delayed ack alarm by
looking at the associated cnt which (where delayed acking is in use)
should probably be 1 if the alarm went off or greater if the ACK was
triggered by a packet.
Signed-off-by: Gavin McCullagh <gavin.mccullagh@nuim.ie> Acked-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Sat, 29 Dec 2007 09:19:49 +0000 (01:19 -0800)]
[SERIAL]: Fix section mismatches in Sun serial console drivers.
We're exporting an __init function, oops :-)
The core issue here is that add_preferred_console() is marked
as __init, this makes it impossible to invoke this thing from
a driver probe routine which is what the Sparc serial drivers
need to do.
There is no harm in dropping the __init marker. This code will
actually work properly when invoked from a modular driver,
except that init will probably not pick up the console change
without some other support code.
Then we can drop the __init from sunserial_console_match()
and we're no longer exporting an __init function to modules.
Signed-off-by: David S. Miller <davem@davemloft.net>
Simon Horman [Fri, 28 Dec 2007 05:19:10 +0000 (21:19 -0800)]
[IPV4] Fix ip=dhcp regression
David Brownell pointed out a regression in my recent "Fix ip command
line processing" patch. It turns out to be a fairly blatant oversight on
my part whereby ic_enable is never set, and thus autoconfiguration is
never enabled. Clearly my testing was broken :-(
The solution that I have is to set ic_enable to 1 if we hit
ip_auto_config_setup(), which basically means that autoconfiguration is
activated unless told otherwise. I then flip ic_enable to 0 if ip=off,
ip=none, ip=::::::off or ip=::::::none using ic_proto_name();
The incremental patch is below, let me know if a non-incremental version
is prepared, as I did as for the original patch to be reverted pending a
fix.
Signed-off-by: Simon Horman <horms@verge.net.au> Signed-off-by: David S. Miller <davem@davemloft.net>
Linus Torvalds [Fri, 28 Dec 2007 05:21:36 +0000 (21:21 -0800)]
[PCI] Do not enable CRS Software Visibility by default
It appears that some PCI-E bridges do the wrong thing in the presense of
CRS Software Visibility and MMCONFIG. In particular, it looks like an
ATI bridge (device ID 7936) will return 0001 in the vendor ID field of
any bridged devices indefinitely.
Not enabling CRS SV avoids the problem, and as we currently do not
really make good use of the feature anyway (we just time out rather than
do any threaded discovery as suggested by the CRS specs), we're better
off just not enabling it.
This should fix a slew of problem reports with random devices (generally
graphics adapters or fairly high-performance networking cards, since it
only affected PCI-E) not getting properly recognized on these AMD systems.
If we really want to use CRS-SV, we may end up eventually needing a
whitelist of systems where this should be enabled, along with some kind
of "pcibios_enable_crs()" query to call the system-specific code.
Suggested-by: Loic Prylli <loic@myri.com> Tested-by: Kai Ruhnau <kai@tragetaschen.dyndns.org> Cc: Matthew Wilcox <matthew@wil.cx> Cc: Greg Kroah-Hartman <greg@kroah.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Geoff Levand [Sun, 23 Dec 2007 17:41:01 +0000 (04:41 +1100)]
[POWERPC] PS3: Fix printing of os-area magic numbers
Fix a bug in the printing of the os-area magic numbers which assumed
that magic numbers were zero terminated strings. The magic numbers
are represented in memory as integers. If the os-area sections are
not initialized correctly they could contained random data that would
be printed to the display. Also unify the handling of header and db
magic numbers and make both of type array of u8.
Signed-off-by: Geoff Levand <geoffrey.levand@am.sony.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
Bob Nelson [Fri, 14 Dec 2007 14:27:30 +0000 (01:27 +1100)]
[POWERPC] Oprofile: Remove dependency on spufs module
This removes an OProfile dependency on the spufs module. This
dependency was causing a problem for multiplatform systems that are
built with support for Oprofile on Cell but try to load the oprofile
module on a non-Cell system.
Signed-off-by: Bob Nelson <rrnelson@us.ibm.com> Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com> Acked-by: Jeremy Kerr <jk@ozlabs.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
Herbert Xu [Fri, 28 Dec 2007 00:05:46 +0000 (11:05 +1100)]
[CRYPTO] padlock: Fix spurious ECB page fault
The xcryptecb instruction always processes an even number of blocks so
we need to ensure th existence of an extra block if we have to process
an odd number of blocks.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Did not fix the reported issue. Apart from other weirdness this causes a
bad link between the TLB flushing logic and the quicklists. If there is
indeed an issue that an arch needs a tlb flush before free then the arch
code needs to set tlb->need_flush before calling quicklist_free.
Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Simon Horman [Wed, 26 Dec 2007 04:54:42 +0000 (20:54 -0800)]
[IPV4]: Fix ip command line processing.
Recently the documentation in Documentation/nfsroot.txt was
update to note that in fact ip=off and ip=::::::off as the
latter is ignored and the default (on) is used.
This was certainly a step in the direction of reducing confusion.
But it seems to me that the code ought to be fixed up so that
ip=::::::off actually turns off ip autoconfiguration.
This patch also notes more specifically that ip=on (aka ip=::::::on)
is the default.
Signed-off-by: Simon Horman <horms@verge.net.au> Signed-off-by: David S. Miller <davem@davemloft.net>
Some users do "modprobe ip_conntrack hashsize=...". Since we have the
module aliases this loads nf_conntrack_ipv4 and nf_conntrack, the
hashsize parameter is unknown for nf_conntrack_ipv4 however and makes
it fail.
Allow to specify hashsize= for both nf_conntrack and nf_conntrack_ipv4.
Note: the nf_conntrack message in the ringbuffer will display an
incorrect hashsize since nf_conntrack is first pulled in as a
dependency and calculates the size itself, then it gets changed
through a call to nf_conntrack_set_hashsize().
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Johannes Berg [Tue, 18 Dec 2007 14:11:35 +0000 (15:11 +0100)]
mac80211: warn when receiving frames with unaligned data
This patch makes mac80211 warn (once) when the driver passes up a
frame in which the payload data is not aligned on a four-byte
boundary, with a long comment for people who run into the condition
and need to know what to do.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
As pointed out by Yanmin Zhang, the problem was already fixed
differently (and correctly), and rather than fix anything, it actually
causes us to create a sub-optimal sched-domains hierarchy (not setting
up the domain belonging to the core) when CONFIG_X86_HT=y.
commit 528a572daea90aa41db92683e5a8756acef514c4 ("ide: add ->chipset field
to ide_pci_device_t") broke hwif->chipset setup (it is now set to ide_cmd646
for CMD648 instead of CMD646). It seems that the breakage happend while
I was moving patches around (cmd64x_chipsets[] entries for CMD646 and CMD648
are identical except for 'name' field). Fix it and bump driver version.
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Use ide_cd_release() to do the cleanup if ide_cdrom_setup() fails.
It fixes:
- the default drive->dsc_overlap value not being restored
- the default drive->queue's prep_rq_fn not being restored
- struct gendisk 'g' not being freed
- wrong function name being reported on unregister_cdrom() error
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
USB: New device ID for the CP2101 driver
USB: VID/PID update for sierra
USB: Unbreak fsl_usb2_udc
Cory T. Tusar [Sun, 23 Dec 2007 20:34:51 +0000 (12:34 -0800)]
tty: fix logic change introduced by wait_event_interruptible_timeout()
Commit 5a52bd4a2dcb570333ce6fe2e16cd311650dbdc8 introduced a subtle logic
change in tty_wait_until_sent(). The original version would only error out
of the 'do { ... } while (timeout)' loop if signal_pending() evaluated to
true; a timeout or break due to an empty buffer would fall out of the loop
and into the tty->driver->wait_until_sent handling. The current
implementation will error out on either a pending signal or an empty
buffer, falling through to the tty->driver->wait_until_sent handling only
on a timeout.
The ->wait_until_sent() will not be reached if the buffer empties before
timeout jiffies have elapsed. This behavior differs from that prior to commit 5a52bd4a2dcb570333ce6fe2e16cd311650dbdc8.
I turned this up while using a little serial download utility to bootstrap an
ARM-based eval board. The util worked fine on 2.6.22.x, but consistently
failed on 2.6.23.x. Once I'd determined that, I narrowed things down with git
bisect, and found the above difference in logic in tty_wait_until_sent() by
inspection.
This change reverts the logic flow in tty_wait_until_sent() to match that
prior to the aforementioned commit.
Signed-off-by: Cory T. Tusar <ctusar@videon-central.com> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Acked-by: Jiri Slaby <jirislaby@gmail.com> Cc: <stable@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Paul Mundt [Sat, 22 Dec 2007 22:03:30 +0000 (14:03 -0800)]
kconfig: obey KCONFIG_ALLCONFIG choices with randconfig.
Currently when using KCONFIG_ALLCONFIG with randconfig the choice options
are clobbered. As recommended by Roman, this adds an is_new test to see
whether to select a new option or obey the existing one.
This is a resend of the earlier patch a couple of weeks ago, since there
was no reply. Original thread is at http://lkml.org/lkml/2007/11/28/94
Signed-off-by: Paul Mundt <lethal@linux-sh.org> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: Sam Ravnborg <sam@ravnborg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Peter Zijlstra [Sat, 22 Dec 2007 22:03:29 +0000 (14:03 -0800)]
lib: proportion: fix underflow in prop_norm_percpu()
Zhe Jiang noticed that its possible to underflow pl->events in
prop_norm_percpu() when the value returned by percpu_counter_read() is less
than the error on that read and the period delay > 1. In that case half might
not trigger the batch increment and the value will be identical on the next
iteration, causing the same half to be subtracted again and again.
Fix this by rewriting the division as a single subtraction instead of a
subtraction loop and using percpu_counter_sum() when the value returned by
percpu_counter_read() is smaller than the error.
The latter is still needed if we want pl->events to shrink properly in the
error region.
[akpm@linux-foundation.org: cleanups] Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Reviewed-by: Jiang Zhe <zhe.jiang@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
projects / mv-sheeva.git / log