I noticed in oprofile study a cache miss in tcp_rcv_established() to read
copied_seq.
ffffffff80400a80 <tcp_rcv_established>: /* tcp_rcv_established total: 4034293
2.0400 */
55493 0.0281 :ffffffff80400bc9: mov 0x4c8(%r12),%eax copied_seq
543103 0.2746 :ffffffff80400bd1: cmp 0x3e0(%r12),%eax rcv_nxt
if (tp->copied_seq == tp->rcv_nxt &&
len - tcp_header_len <= tp->ucopy.len) {
In this function, the cache line 0x4c0 -> 0x500 is used only for this
reading 'copied_seq' field.
rcv_wup and copied_seq should be next to rcv_nxt field, to lower number of
active cache lines in hot paths. (tcp_rcv_established(), tcp_poll(), ...)
As you suggested, I changed tcp_create_openreq_child() so that these fields
are changed together, to avoid adding a new store buffer stall.
Patch is 64bit friendly (no new hole because of alignment constraints)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This interpretation comes from RFC4138:
"If the sender implements some loss recovery algorithm other
than Reno or NewReno [FHG04], the F-RTO algorithm SHOULD
NOT be entered when earlier fast recovery is underway."
I think the RFC means to say (especially in the light of
Appendix B) that ...recovery is underway (not just fast recovery)
or was underway when it was interrupted by an earlier (F-)RTO
that hasn't yet been resolved (snd_una has not advanced enough).
Thus, my interpretation is that whenever TCP has ever
retransmitted other than head, basic version cannot be used
because then the order assumptions which are used as FRTO basis
do not hold.
NewReno has only the head segment retransmitted at a time.
Therefore, walk up to the segment that has not been SACKed, if
that segment is not retransmitted nor anything before it, we know
for sure, that nothing after the non-SACKed segment should be
either. This assumption is valid because TCPCB_EVER_RETRANS does
not leave holes but each non-SACKed segment is rexmitted
in-order.
Check for retrans_out > 1 avoids more expensive walk through the
skb list, as we can know the result beforehand: F-RTO will not be
allowed.
SACKed skb can turn into non-SACked only in the extremely rare
case of SACK reneging, in this case we might fail to detect
retransmissions if there were them for any other than head. To
get rid of that feature, whole rexmit queue would have to be
walked (always) or FRTO should be prevented when SACK reneging
happens. Of course RTO should still trigger after reneging which
makes this issue even less likely to show up. And as long as the
response is as conservative as it's now, nothing bad happens even
then.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
A security issue is emerging. Disallow Routing Header Type 0 by default
as we have been doing for IPv4.
Note: We allow RH2 by default because it is harmless.
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get rid of the inlined #ifdefs.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I've traced down an off-by-one TCP checksum calculation error under
the following conditions:
1) The TCP code needs to split a full-sized packet due to a reduced
MSS (typically due to the addition of TCP options mid-stream like
SACK).
_AND_
2) The checksum of the 2nd fragment is larger than the checksum of the
original packet. After subtraction this results in a checksum for
the 1st fragment with bits 16..31 set to 1. (this is ok)
_AND_
3) The checksum of the 1st fragment's TCP header plus the previously
32bit checksum of the 1st fragment DOES NOT cause a 32bit overflow
when added together. This results in a checksum of the TCP header
plus TCP data that still has the upper 16 bits as 1's.
_THEN_
4) The TCP+data checksum is added to the checksum of the pseudo IP
header with csum_tcpudp_nofold() incorrectly (the bug).
The problem is the checksum of the TCP+data is passed to
csum_tcpudp_nofold() as an 32bit unsigned value, however the assembly
code acts on it as if it is a 64bit unsigned value.
This causes an incorrect 32->64bit extension if the sum has bit 31
set. The resulting checksum is off by one.
This problems is data and TCP header dependent due to #2 and #3
above so it doesn't occur on every TCP packet split.
Signed-off-by: Dave Johnson <djohnson+linux-mips@sw.starentnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
With commit 63dc68a8cf, kernel can not
handle BUG() and BUG_ON() properly since get_user() returns false for
kernel code. Use __get_user() to skip unnecessary access_ok(). This
patch also make BRK_BUG code encoded in the TNE instruction.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The save_fp_context()/restore_fp_context() might sleep on accessing
user stack and therefore might lose FPU ownership in middle of them.
If these function failed due to "in_atomic" test in do_page_fault,
touch the sigcontext area in non-atomic context and retry these
save/restore operation.
This is a replacement of a (broken) fix which was titled "Allow CpU
exception in kernel partially".
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The commit 4d40bff7110e9e1a97ff8c01bdd6350e9867cc10 ("Allow CpU
exception in kernel partially") was broken. The commit was to fix
theoretical problem but broke usual case. Revert it for now.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Recent versions of the BCM112X processors aren't recognized by Linux
(preventing Linux from booting on those processors). This patch adds
support for those that are missing.
Signed-off-by: Mark Mason <mason@broadcom.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
[BRIDGE]: Unaligned access when comparing ethernet addresses
[SCTP]: Unmap v4mapped addresses during SCTP_BINDX_REM_ADDR operation.
[SCTP]: Fix assertion (!atomic_read(&sk->sk_rmem_alloc)) failed message
[NET]: Set a separate lockdep class for neighbour table's proxy_queue
[NET]: Fix UDP checksum issue in net poll mode.
[KEY]: Fix conversion between IPSEC_MODE_xxx and XFRM_MODE_xxx.
[NET]: Get rid of alloc_skb_from_cache
Provide an dummy implementation of devm_ioport_map() and
devm_ioport_unmap() to allow drivers (eg, pata_platform) to build for
platforms where CONFIG_NO_IOPORT is selected.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Files:
arch/alpha/kernel/core_mcpcia.c
arch/alpha/kernel/sys_rawhide.c
include/asm-alpha/core_mcpcia.h
Determine correct hose configuration; RAWHIDE family can have
2 or 4 hoses, so make sure non-existent hoses are ignored.
arch/alpha/kernel/err_titan.c
Supply a needed #include <asm/irq_regs.h>
arch/alpha/kernel/module.c
Add some useful output to the relocation overflow messages.
arch/alpha/kernel/sys_noritake.c
Supply necessary noritake_end_irq() to correct interrupt handling.
This fixes a problem first noted by hangs during boot probing with
a DE500-BA TULIP NIC present.
arch/alpha/kernel/sys_sio.c
Correct saving of original PIRQ register (PCI IRQ routing);
change default PIRQ setting to leave PCI IRQs 9 and 14 free to
be used for sound (Multia) and IDE (any), respectively.
include/asm-alpha/io.h
Supply the "isa_virt_to_bus" routine.
Signed-off-by: Jay Estabrook <jay.estabrook@hp.com>
Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Make kernel-doc comments match macro names.
Correct parameter names in a few places.
Remove '#' from beginning of kernel-doc comment macro names.
Remove extra (erroneous) blank lines in kernel-doc.
Warning(plist.h:100): Cannot understand * #PLIST_HEAD_INIT - static struct plist_head initializer on line 100 - I thought it was a doc line
Warning(plist.h:112): Cannot understand * #PLIST_NODE_INIT - static struct plist_node initializer on line 112 - I thought it was a doc line
Warning(plist.h:103): No description found for parameter '_lock'
Warning(plist.h:129): No description found for parameter 'lock'
Warning(plist.h:158): No description found for parameter 'pos'
Warning(plist.h:169): No description found for parameter 'pos'
Warning(plist.h:169): No description found for parameter 'n'
Warning(plist.h:179): No description found for parameter 'mem'
This still leaves one warning & one error that need attention:
Error(plist.h:219): cannot understand prototype: '('
Warning(plist.h): no structured comments found
Acked-by: Inaky Perez-Gonzalez <inaky.perez-gonzalez@intel.com>
Cc: Daniel Walker <dwalker@mvista.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Trivial change to pass vmsplice arguments through the compat layer on
pp64.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Acked-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Otherwise the following calltrace will lead to a wrong
lockdep warning:
neigh_proxy_process()
`- lock(neigh_table->proxy_queue.lock);
arp_redo /* via tbl->proxy_redo */
arp_process
neigh_event_ns
neigh_update
skb_queue_purge
`- lock(neighbor->arp_queue.lock);
This is not a deadlock actually, as neighbor table's proxy_queue
and the neighbor's arp_queue are different queues.
Lockdep thinks there is a deadlock as both queues are initialized
with skb_queue_head_init() and thus have a common class.
Signed-off-by: David S. Miller <davem@davemloft.net>
Since this was added originally for Xen, and Xen has recently (~2.6.18)
stopped using this function, we can safely get rid of it. Good timing
too since this function has started to bit rot.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the writebacks are cancelled via nfs_cancel_dirty_list, or due to the
memory allocation failing in nfs_flush_one/nfs_flush_multi, then we must
ensure that the PG_writeback flag is cleared.
Also ensure that we actually own the PG_writeback flag whenever we
schedule a new writeback by making nfs_set_page_writeback() return the
value of test_set_page_writeback().
The PG_writeback page flag ends up replacing the functionality of the
PG_FLUSHING nfs_page flag, so we rip that out too.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* master.kernel.org:/pub/scm/linux/kernel/git/bart/ide-2.6:
ide: add "optical" to sysfs "media" attribute
ide: ugly messages trying to open CD drive with no media present
ide: correctly prevent IDE timer expiry function to run if request was already handled
It is possible for the timer expiry function to run even though the
request has already been handled: ide_timer_expiry() only checks that
the handler is not NULL, but it is possible that we have handled a
request (thus clearing the handler) and then started a new request
(thus starting the timer again, and setting a handler).
A simple way to exhibit this is to set the DMA timeout to 1 jiffy and
run dd: The kernel will panic after a few minutes because
ide_timer_expiry() tries to add a timer when it's already active.
To fix this, we simply add a request generation count that gets
incremented at every interrupt, and check in ide_timer_expiry() that
we have not already handled a new interrupt before running the expiry
function.
Signed-off-by: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Since lazy MMU batching mode still allows interrupts to enter, it is
possible for interrupt handlers to try to use kmap_atomic, which fails when
lazy mode is active, since the PTE update to highmem will be delayed. The
best workaround is to issue an explicit flush in kmap_atomic_functions
case; this is the only way nested PTE updates can happen in the interrupt
handler.
Thanks to Jeremy Fitzhardinge for noting the bug and suggestions on a fix.
This patch gets reverted again when we start 2.6.22 and the bug gets fixed
differently.
Signed-off-by: Zachary Amsden <zach@vmware.com>
Cc: Andi Kleen <ak@muc.de>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Soeren Sonnenburg reported that upon resume he is getting
this backtrace:
[<c0119637>] smp_apic_timer_interrupt+0x57/0x90
[<c0142d30>] retrigger_next_event+0x0/0xb0
[<c0104d30>] apic_timer_interrupt+0x28/0x30
[<c0142d30>] retrigger_next_event+0x0/0xb0
[<c0140068>] __kfifo_put+0x8/0x90
[<c0130fe5>] on_each_cpu+0x35/0x60
[<c0143538>] clock_was_set+0x18/0x20
[<c0135cdc>] timekeeping_resume+0x7c/0xa0
[<c02aabe1>] __sysdev_resume+0x11/0x80
[<c02ab0c7>] sysdev_resume+0x47/0x80
[<c02b0b05>] device_power_up+0x5/0x10
it turns out that on resume we mistakenly re-enable interrupts too
early. Do the timer retrigger only on the current CPU.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Soeren Sonnenburg <kernel@nn7.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
On a SGI Altix TIOCP based PCI bus we need to include the ATE_PIO attribute
bit if we're mapping a 32bit MSI address.
Signed-off-by: Mike Habeck <habeck@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Just a one-byter for an ia64 thinko/typo - already fixed for i386 and x86_64.
Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Revert all this. It can cause device-mapper to receive a different major from
earlier kernels and it turns out that the Amanda backup program (via GNU tar,
apparently) checks major numbers on files when performing incremental backups.
Which is a bit broken of Amanda (or tar), but this feature isn't important
enough to justify the churn.
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A device can be removed from an md array via e.g.
echo remove > /sys/block/md3/md/dev-sde/state
This will try to remove the 'dev-sde' subtree which will deadlock
since
commit e7b0d26a86
With this patch we run the kobject_del via schedule_work so as to
avoid the deadlock.
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
patch 4/4:
Limit ATAPI DMA to R/W commands only for TORiSAN DRD-N216 DVD-ROM drives
(http://bugzilla.kernel.org/show_bug.cgi?id=6710)
Signed-off-by: Albert Lee <albertcc@tw.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
patch 3/4:
The TORiSAN drive locks up when max sector == 256.
Limit max sector to 128 for the TORiSAN DRD-N216 drives.
(http://bugzilla.kernel.org/show_bug.cgi?id=6710)
Signed-off-by: Albert Lee <albertcc@tw.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
patch 1/4:
Reorder HSM_ST_FIRST, such that the task state transition is easier decoded with human eyes.
Signed-off-by: Albert Lee <albertcc@tw.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Add unsigned to unused bit field in a.out.h to make sparse happy.
[ I took care of the sparc64 side as well -DaveM ]
Signed-off-by: Robert Reif <reif@earthlink.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6:
[PATCH] x86: Don't probe for DDC on VBE1.2
[PATCH] x86-64: Increase NMI watchdog probing timeout
[PATCH] x86-64: Let oprofile reserve MSR on all CPUs
[PATCH] x86-64: Disable local APIC timer use on AMD systems with C1E
Fix the regression resulting from the recent change of suspend code
ordering that causes systems based on Intel x86 CPUs using the microcode
driver to hang during the resume.
The problem occurs since the microcode driver uses request_firmware() in
its CPU hotplug notifier, which is called after tasks has been frozen and
hangs. It can be fixed by telling the microcode driver to use the
microcode stored in memory during the resume instead of trying to load it
from disk.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Adrian Bunk <bunk@stusta.de>
Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Maxim <maximlevitsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently we have a confused udelay implementation.
* __const_udelay does not accept usecs but xloops in i386 and x86_64
* our implementation requires usecs as arg
* it gets a xloops count when called by asm/arch/delay.h
Bugs related to this (extremely long shutdown times) where reported by some
x86_64 users, especially using Device Mapper.
To hit this bug, a compile-time constant time parameter must be passed -
that's why UML seems to work most times. Fix this with a simple udelay
implementation.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
AMD dual core laptops with C1E do not run the APIC timer correctly
when they go idle. Previously the code assumed this only happened
on C2 or deeper. But not all of these systems report support C2.
Use a AMD supplied snippet to detect C1E being enabled and then disable
local apic timer use.
This supercedes an earlier workaround using DMI detection of specific systems.
Thanks to Mark Langsdorf for the detection snippet.
Signed-off-by: Andi Kleen <ak@suse.de>
This patch:
- Switches mb/rmb/wmb back to being full-blown DMBs on ARM SMP systems,
since mb/rmb/wmb are required to order Normal memory accesses as well.
- Enables the use of DMB and ISB on XSC3 (which is an ARMv5TE ISA core
but conforms to the ARMv6 memory ordering model and supports the
various ARMv6 barriers.)
- Makes DMA coherent platforms (only ixp23xx at the moment) map
mb/rmb/wmb to dmb(), as on DMA coherent platforms, DMA consistent
mappings are done as Normal mappings, which are weakly ordered.
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Acked-by: David Howells <dhowells@redhat.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
There was a typo in commit 7632fc8f80,
preventing it from working - 32bit binaries crashed hopelessly before
the below fix and work perfectly now.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix compilation fail for ixp4xx platforms for the case when CONFIG_IXP4XX_INDIRECT_PCI is set. That is due to the check_signature() is appeared in include/linux/io.h.
Signed-off-by: Vladimir Barinov <vbarinov@ru.mvista.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
local_irq_restore -> raw_local_irq_restore -> irq_restore_epilog ->
smtc_ipi_replay -> smtc_ipi_dq -> spin_unlock_irqrestore ->
_spin_unlock_irqrestore -> local_irq_restore
The recursion does abort when there is no more IPI queued for a CPU, so
this isn't usually fatal which is why we got away with this for so long
until this was discovered by code inspection.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
