Martin Peschke [Thu, 27 Mar 2008 13:22:02 +0000 (14:22 +0100)]
[SCSI] zfcp: Add traces for state changes.
This patch writes a trace record which provides information about state
changes for adapters, ports and units, e.g. target failure, targets becoming
online, targets being temporarily blocked due to pending recovery, targets
which have been recovered successfully etc.
Signed-off-by: Martin Peschke <mp3@de.ibm.com> Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Martin Peschke [Thu, 27 Mar 2008 13:21:58 +0000 (14:21 +0100)]
[SCSI] zfcp: Add qtcb dump to hba debug trace
This patch adds per request hardware debugging data to the trace
record which is written per request. It's a replacement for some sad
kernel message based debugging code. Considering the amount of trace
data, printk() is not suitable for this stuff. Writing binary traces
is more efficient. In addition we got all information in one place.
The QTCB trace data is only dumped for requests other than SCSI
requests. Otherwise we would flood the trace ring buffer. We are
mostly interested in non-SCSI, recovery related requests here anyway.
This patch also works around a known hardware bug. It truncates QTCB
traces so that we do not save unused areas of the hardware trace.
Signed-off-by: Martin Peschke <mp3@de.ibm.com> Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
James Bottomley [Sun, 30 Mar 2008 16:38:07 +0000 (11:38 -0500)]
[SCSI] mpt fusion: fix up msi_enable in mpt_suspend
There's a problem with the combination of the upstream power
management fixes and the enabling of MSI by default in that the
suspend path still uses the global variable. Convert it to check
ioc->msi_enable.
FUJITA Tomonori [Tue, 25 Mar 2008 00:26:52 +0000 (09:26 +0900)]
[SCSI] stex: use scsi_build_sense_buffer
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Acked-by: Ed Lin <ed.lin@promise.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
[SCSI] WD33C93: let platform stub override no_sync/fast/dma_mode
SGI machines with WD33C93 allow usage of burst mode DMA, which increases
performance noticable. To make this selectable by the sgiwd93 stub,
setting the values for no_sync, fast and dma_mode has been moved to the
individual platform stubs.
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Acked-by: Ralf Baechle <ralf@linux-mips.org> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Michael Reed [Thu, 20 Mar 2008 22:32:05 +0000 (17:32 -0500)]
[SCSI] mptsas: do not use ioc->handle to locate hba portinfo structure
While performing hardware raid reset testing via the raid's client, I
noticed that sometimes, following the reset, that there would be more
raid targets in the lsscsi output than there actually were raid
targets. I tracked this down to the following issue.
Fusion cannot always find the mptsas_portinfo structure for the hba
because it uses the handle stored in ioc->handle to locate it. The
problem is that the firmware can change the handle associated with the
hba when h/w raid is reset (via the raid client). When this happens,
the driver will allocate another mptsas_portinfo structure and link it
into the chain of said structures. This ultimately causes confusion
within the driver resulting in targets not being removed when they
should be.
Eric Moore pointed out that the hba's portinfo structure is always the
first structure on the sas_topology list. This patch modifies
mptsas.c to access the hba's portinfo structure by taking the first
structure on said list.
Signed-off-by: Michael Reed <mdr@sgi.com> Acked-by: "Moore, Eric" <Eric.Moore@lsi.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
1. when hba completion status is good, check for iscsi transport
errors (underflow/overflow) prior to checking the scsi status
2. New firmware requires that one marker iocb be issued for each task
management command. The patch issues marker iocb immediately
following a LUN or Target reset.
Signed-off-by: David C Somayajulu <david.somayajulu@qlogic.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
bo yang [Mon, 17 Mar 2008 07:36:43 +0000 (03:36 -0400)]
[SCSI] megaraid_sas: Fix the frame count calculation
When Driver sent wrong frame count to firmware. As this particular
command is sent to drive, FW is seeing continuous chip resets and so
the command will timeout.
Signed-off-by Bo Yang<bo.yang@lsi.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Roel Kluin [Mon, 28 Jan 2008 13:09:54 +0000 (14:09 +0100)]
[SCSI] aic7xxx: Test opcode, not definition in aicasm:type_check()
This fixes a bug that we treat all sequencer operations as ands and
never do the additional invalid bit checks non-and operations require
because the if () to determine this has an operand which is always
true at the end of the or statement.
Signed-off-by: Roel Kluin <12o3l@tiscali.nl> Acked-by: Hannes Reinecke <hare@suse.de> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
FUJITA Tomonori [Thu, 20 Mar 2008 02:09:19 +0000 (11:09 +0900)]
[SCSI] scsi_debug: remove unnecessary function declarations
This patch removes lots of function declarations with moving
scsi_debug_queuecommand. This cleans up scsi_debug_queuecommand a bit
to silence checkpatch.pl
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Acked-by: Douglas Gilbert <dougg@torque.net> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
FUJITA Tomonori [Thu, 20 Mar 2008 02:09:16 +0000 (11:09 +0900)]
[SCSI] scsi_debug: create new scsi_debug devices at a single place
Two functions, sdebug_add_adapter and devInfoReg, creates new
scsi_debug devices. To simplify the code, this patch adds a new helper
function to create new scsi_debug devices (sdebug_device_create) and
converts both functions to use it.
I plan to add more to scsi_debug devices (e.g. using a thread for a
scsi_debug device for scalability testings). This patch enable me to
add such to just the new helper function instead of touching two
functions, sdebug_add_adapter and devInfoReg.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Acked-by: Douglas Gilbert <dougg@torque.net> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
James Bottomley [Thu, 13 Mar 2008 16:16:33 +0000 (11:16 -0500)]
[SCSI] consolidate command allocation in a single place
Since the way we allocate commands with a separate sense buffer is
getting complicated, we should isolate setup and teardown to a single
routine so that if it gets even more complex, there's only one place
in the code that needs to be altered.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
FUJITA Tomonori [Sun, 9 Mar 2008 04:44:35 +0000 (13:44 +0900)]
[SCSI] stex: use sg buffer copy helper functions
This replaces stex_internal_copy with scsi_sg_copy_to/from_buffer.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Acked-by: Ed Lin <ed.lin@promise.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
FUJITA Tomonori [Sun, 9 Mar 2008 04:44:31 +0000 (13:44 +0900)]
[SCSI] ps3rom: use sg buffer copy helper funcitons
Note that if scsi_bufflen(cmd) is not zero, the command always has an
sg list. So this patch doesn't do the error checking in
fill_from_dev_buffer and fetch_to_dev_buffer did.
FUJITA Tomonori [Mon, 17 Mar 2008 15:15:03 +0000 (00:15 +0900)]
[SCSI] block: add sg buffer copy helper functions
This patch adds new three helper functions to copy data between an SG
list and a linear buffer.
- sg_copy_from_buffer copies data from linear buffer to an SG list
- sg_copy_to_buffer copies data from an SG list to a linear buffer
When the APIs copy data from a linear buffer to an SG list,
flush_kernel_dcache_page is called. It's not necessary for everyone
but it's a no-op on most architectures and in general the API is not
used in performance critical path.
Matthew Wilcox [Thu, 13 Mar 2008 12:55:08 +0000 (06:55 -0600)]
[SCSI] BusLogic: make FlashPoint support x86-32 only
We've verified that there are 64 bit and endianness problems in the
flashpoint driver. Reverse the logic of CONFIG_OMIT_FLASHPOINT (make
it CONFIG_SCSI_FLASHPOINT) and make it depend on X86_32 so it can't
appear for any other architectures. Long term, if someone chooses,
they could make FlashPoint 64 bit compliant (it looks like its a
question of fixing up the sizes in some of the packed descriptors)
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Christof Schmitt [Mon, 10 Mar 2008 15:18:54 +0000 (16:18 +0100)]
[SCSI] zfcp: Fix handling for boxed port after physical close
When a FSF physical close returns the status boxed, this means that
another system already closed the port. For our system this is the
same status as in the good path, we have to send the normal close. So,
set the status for the boxed response to the same as for the good
status.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com> Signed-off-by: Martin Peschke <mp3@de.ibm.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
[SCSI] zfcp: convert zfcp to use target reset and device reset handler
[based on proposal from Mike Christie <michaelc@cs.wisc.edu>, this
patch adds some simplifications to the handler functions]
With the new target reset handler callback in the SCSI midlayer, the
device reset handler in zfcp can be split in two parts. Now, zfcp does
not have to track anymore whether the device supports LUN resets, so
remove this flag and let the SCSI midlayer decide what to do.
The device reset handler simply issues a LUN reset and the target
reset handler a target reset.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Mike Christie [Sat, 1 Mar 2008 00:25:20 +0000 (18:25 -0600)]
[SCSI] qla4xxx: Add target reset functionality
This patch adds target reset functionalty.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Acked-by: David Somayajulu <david.somayajulu@qlogic.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Mike Christie [Sat, 1 Mar 2008 00:25:19 +0000 (18:25 -0600)]
[SCSI] scsi_error: add target reset handler
The problem is that serveral drivers are sending a target reset from the
device reset handler, and if we have multiple devices a target reset gets
sent for each device when only one would be sufficient. And if we do a target
reset it affects all the commands on the target so the device reset handler
code only cleaning up one devices's commands makes programming the driver a
little more difficult than it should be.
This patch adds a target reset handler, which drivers can use to send
a target reset. If successful it cleans up the commands for a devices
accessed through that starget.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
As we no longer need to calculate the data length of the whole scatterlist,
we can abort the loop earlier and coalesce req_len and act_len into one
variable, making fill_from_dev_buffer() more similar to fetch_to_dev_buffer().
Signed-off-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
FUJITA Tomonori [Sun, 24 Feb 2008 23:25:23 +0000 (15:25 -0800)]
[SCSI] aacraid: READ_CAPACITY_16 shouldn't trust allocation length in cdb
When aacraid spoofs READ_CAPACITY_16, it assumes that the data length
in the sg list is equal to allocation length in cdb. But sg can put
any value in scb so the driver needs to check both the data length in
the sg list and allocation length in cdb.
If allocation length is larger than the response length that the
driver expects, it clears the data buffer in the sg list to zero but
it doesn't need to do. Just setting resid is fine.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Acked-by: Mark Salyzyn <Mark_Salyzyn@adaptec.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Kai Makisara [Sun, 24 Feb 2008 20:29:12 +0000 (22:29 +0200)]
[SCSI] st: show options currently set in sysfs
Show the current binary tape driver and mode options is sysfs. A file
(options) is created in each directory in /sys/class/scsi_tape. The files
contain masks showing the options. The mask bit definitions are the same as
used when setting the options using the MTSETDRVBUFFER function in the
MTIOCTOP ioctl (defined in include/linux/mtio.h). For example:
> cat /sys/class/scsi_tape/nst0/options
0x00000d07
[jejb: updated doc with correction from Randy Dunlap]
Signed-off-by: Kai Makisara <kai.makisara@kolumbus.fi> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Kai Makisara [Sun, 24 Feb 2008 20:23:24 +0000 (22:23 +0200)]
[SCSI] st: add option to use SILI in variable block reads
Add new option MT_ST_SILI to enable setting the SILI bit in reads in variable
block mode. If SILI is set, reading a block shorter than the byte count does
not result in CHECK CONDITION. The length of the block is determined using the
residual count from the HBA. Avoiding the REQUEST SENSE command for every
block speeds up some real applications considerably.
Signed-off-by: Kai Makisara <kai.makisara@kolumbus.fi> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Darrick J. Wong [Tue, 19 Feb 2008 18:49:40 +0000 (10:49 -0800)]
[SCSI] libsas: Provide a transport-level facility to request SAS addrs
Provide a facility to use the request_firmware() interface to get a SAS
address from userspace. This can be used by SAS LLDDs that cannot
obtain the address from the host adapter.
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
FUJITA Tomonori [Tue, 19 Feb 2008 09:41:30 +0000 (18:41 +0900)]
[SCSI] ips: sg chaining support to the path to non I/O commands
I overlooked ips_scmd_buf_write and ips_scmd_buf_read when I converted
ips to use the data buffer accessors.
ips is unlikely to use sg chaining (especially in this path) since a)
this path is used only for non I/O commands (with little data
transfer), b) ips's sg_tablesize is set to just 17.
Thanks to Tim Pepper for testing this patch.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Acked-by: Mark Salyzyn <Mark_Salyzyn@adaptec.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Jeff Garzik [Sat, 16 Feb 2008 03:20:09 +0000 (22:20 -0500)]
[SCSI] gdth: convert to PCI hotplug API
- remove PCI device sort, which greatly simplifies PCI probe,
permitting direct, per-HBA function calls rather than an indirect
route to the same end result.
- remove need for pcistr[]
Signed-off-by: Jeff Garzik <jgarzik@redhat.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
fix endian lossage in forcedeth
net/tokenring/olympic.c section fixes
net: marvell.c fix sparse shadowed variable warning
[VLAN]: Fix egress priority mappings leak.
[TG3]: Add PHY workaround for 5784
[NET]: srandom32 fixes for networking v2
[IPV6]: Fix refcounting for anycast dst entries.
[IPV6]: inet6_dev on loopback should be kept until namespace stop.
[IPV6]: Event type in addrconf_ifdown is mis-used.
[ICMP]: Ensure that ICMP relookup maintains status quo
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
[SPARC64]: Fix user accesses in regset code.
[SPARC64]: Fix FPU saving in 64-bit signal handling.
Merge branch 'pci_id_updates' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/v4l-dvb
* 'pci_id_updates' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/v4l-dvb:
V4L/DVB (7497): pvrusb2: add new usb pid for 73xxx models
V4L/DVB (7496): pvrusb2: add new usb pid for 75xxx models
Rusty Russell [Sun, 6 Apr 2008 07:23:38 +0000 (17:23 +1000)]
Fix booting pentium+ with dodgy TSC
We handle a broken tsc these days, so no need to panic. We clear the
TSC bit when tsc_init decides it's unreliable (eg. under lguest w/ bad
host TSC), leading to bogus panic.
Now that we're mapping registers in the DRM driver at load time, the
driver actually checks the PCI ID, so we need to make sure the macros
have all the right bits (and longer term use the DRM headers as the sole
copy of the PCI & register definitions).
This patch adds 945GME support to the DRM headers, fixing a regression
reported in http://bugzilla.kernel.org/show_bug.cgi?id=10395.
Since 2.6.25-rc7, I've been seeing an occasional livelock on one x86_64
machine, copying kernel trees to tmpfs, paging out to swap.
Signature: 6000 pages under writeback but never getting written; most
tasks of interest trying to reclaim, but each get_swap_bio waiting for a
bio in mempool_alloc's io_schedule_timeout(5*HZ); every five seconds an
atomic page allocation failure report from kblockd failing to allocate a
sense_buffer in __scsi_get_command.
__scsi_get_command has a (one item) free_list to protect against this,
but rc1's [SCSI] use dynamically allocated sense buffer de25deb18016f66dcdede165d07654559bb332bc upset that slightly. When it
fails to allocate from the separate sense_slab, instead of giving up, it
must fall back to the command free_list, which is sure to have a
sense_buffer attached.
Either my earlier -rc testing missed this, or there's some recent
contributory factor. One very significant factor is SLUB, which merges
slab caches when it can, and on 64-bit happens to merge both bio cache
and sense_slab cache into kmalloc's 128-byte cache: so that under this
swapping load, bios above are liable to gobble up all the slots needed
for scsi_cmnd sense_buffers below.
That's disturbing behaviour, and I tried a few things to fix it. Adding
a no-op constructor to the sense_slab inhibits SLUB from merging it, and
stops all the allocation failures I was seeing; but it's rather a hack,
and perhaps in different configurations we have other caches on the
swapout path which are ill-merged.
Another alternative is to revert the separate sense_slab, using
cache-line-aligned sense_buffer allocated beyond scsi_cmnd from the one
kmem_cache; but that might waste more memory, and is only a way of
diverting around the known problem.
While I don't like seeing the allocation failures, and hate the idea of
all those bios piled up above a scsi host working one by one, it does
seem to emerge fairly soon with the livelock fix. So lacking better
ideas, stick with that one clear fix for now.
Signed-off-by: Hugh Dickins <hugh@veritas.com> Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Christoph Lameter <clameter@sgi.com> Cc: Pekka Enberg <penberg@cs.helsinki.fi> Cc: Peter Zijlstra <a.p.ziljstra@chello.nl> Cc: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Andrew Paprocki [Wed, 2 Apr 2008 06:43:19 +0000 (02:43 -0400)]
[WATCHDOG] it8712f_wdt Zero MSB timeout byte when disabling watchdog
I noticed this while testing the latest code. I'm not sure if it is required,
but the normal (or LSB) timeout value is set to zero, so the MSB should
be as well to stay consistent.
If the chip revision is >= 8, set MSB of the 16-bit timeout value to zero
when disabling the watchdog in it8712f_wdt_disable().
Signed-off-by: Andrew Paprocki <andrew@ishiboo.com> Signed-off-by: Wim Van Sebroeck <wim@iguana.be> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Revert "ACPI: Ignore _BQC object when registering backlight device"
This reverts commit 7c0ea45be4f114d85ee35caeead8e1660699c46f which
caused a regression with the backlight being set to off when a laptop
doesn't have a _BQC entry to query the actual backlight value. The code
blindly then falls back on a value of 0.
See
http://bugzilla.kernel.org/show_bug.cgi?id=10387
http://lkml.org/lkml/2008/4/2/366
for details.
Bisected-and-reported-by: Andrey Borzenkov <arvidjaar@mail.ru> Cc: Zhao Yakui <yakui.zhao@intel.com> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Len Brown <len.brown@intel.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Carol Hebert [Fri, 4 Apr 2008 21:30:03 +0000 (14:30 -0700)]
ipmi: change device node ordering to reflect probe order
In 2.6.14 a patch was merged which switching the order of the ipmi device
naming from in-order-of-discovery over to reverse-order-of-discovery.
So on systems with multiple BMC interfaces, the ipmi device names are being
created in reverse order relative to how they are discovered on the system
(e.g. on an IBM x3950 multinode server with N nodes, the device name for the
BMC in the first node is /dev/ipmiN-1 and the device name for the BMC in the
last node is /dev/ipmi0, etc.).
The problem is caused by the list handling routines chosen in dmi_scan.c.
Using list_add() causes the multiple ipmi devices to be added to the device
list using a stack-paradigm and so the ipmi driver subsequently pulls them off
during initialization in LIFO order. This patch changes the
dmi_save_ipmi_device() list handling paradigm to a queue, thereby allowing the
ipmi driver to build the ipmi device names in the order in which they are
found on the system.
Signed-off-by: Carol Hebert <cah@us.ibm.com> Signed-off-by: Corey Minyard <cminyard@mvista.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mtd: fix broken state in CFI driver caused by FL_SHUTDOWN
THe CFI driver in 2.6.24 kernel is broken. Not so intensive read/write
operations cause incomplete writes which lead to kernel panics in JFFS2.
We investigated the issue - it is caused by bug in FL_SHUTDOWN parsing code.
Sometimes chip returns -EIO as if it is in FL_SHUTDOWN state when it should
wait in FL_PONT (error in order of conditions).
The following patch fixes the bug in state parsing code of CFI. Also I've
added comments to notify developers if they want to add new case in future.
Signed-off-by: Alexey Korolev <akorolev@infradead.org> Reviewed-by: Joern Engel <joern@logfs.org> Cc: David Woodhouse <dwmw2@infradead.org> Cc: <stable@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
memory controller: make memory resource control aware of boot options
A boot option for the memory controller was discussed on lkml. It is a good
idea to add it, since it saves memory for people who want to turn off the
memory controller.
By default the option is on for the following two reasons:
1. It provides compatibility with the current scheme where the memory
controller turns on if the config option is enabled
2. It allows for wider testing of the memory controller, once the config
option is enabled
We still allow the create, destroy callbacks to succeed, since they are not
aware of boot options. We do not populate the directory will memory resource
controller specific files.
Paul Menage [Fri, 4 Apr 2008 21:29:57 +0000 (14:29 -0700)]
cgroups: add cgroup support for enabling controllers at boot time
The effects of cgroup_disable=foo are:
- foo isn't auto-mounted if you mount all cgroups in a single hierarchy
- foo isn't visible as an individually mountable subsystem
As a result there will only ever be one call to foo->create(), at init time;
all processes will stay in this group, and the group will never be mounted on
a visible hierarchy. Any additional effects (e.g. not allocating metadata)
are up to the foo subsystem.
This doesn't handle early_init subsystems (their "disabled" bit isn't set be,
but it could easily be extended to do so if any of the early_init systems
wanted it - I think it would just involve some nastier parameter processing
since it would occur before the command-line argument parser had been run.
Hugh said:
Ballpark figures, I'm trying to get this question out rather than
processing the exact numbers: CONFIG_CGROUP_MEM_RES_CTLR adds 15% overhead
to the affected paths, booting with cgroup_disable=memory cuts that back to
1% overhead (due to slightly bigger struct page).
I'm no expert on distros, they may have no interest whatever in
CONFIG_CGROUP_MEM_RES_CTLR=y; and the rest of us can easily build with or
without it, or apply the cgroup_disable=memory patches.
Unix bench's execl test result on x86_64 was
== just after boot without mounting any cgroup fs.==
mem_cgorup=off : Execl Throughput 43.0 3150.1 732.6
mem_cgroup=on : Execl Throughput 43.0 2932.6 682.0
==
Sergei Shtylyov [Thu, 20 Mar 2008 17:59:34 +0000 (20:59 +0300)]
[MIPS] Make KGDB compile on UP
Building UP kernel with KGDB enabled produces the following errors and warning
(fatal due to -Werror in arch/mips/kernel/Makefile):
In file included from arch/mips/kernel/gdb-stub.c:142:
include/asm/smp.h:25:1: "raw_smp_processor_id" redefined
In file included from include/linux/sched.h:69,
from arch/mips/kernel/gdb-stub.c:126:
include/linux/smp.h:88:1: this is the location of the previous definition
In file included from arch/mips/kernel/gdb-stub.c:142:
include/asm/smp.h:62: error: redefinition of 'smp_send_reschedule'
include/linux/smp.h:102: error: previous definition of 'smp_send_reschedule' was here
include/asm/smp.h: In function `smp_send_reschedule':
include/asm/smp.h:65: error: dereferencing pointer to incomplete type
arch/mips/kernel/gdb-stub.c: At top level:
arch/mips/kernel/gdb-stub.c:660: warning: 'kgdb_wait' defined but not used
Fix the errors by not directly including <asm/smp.h> (which is already included
by <linux/smp.h>) and the warning by enclosing kgdb_wait() in #ifdef CONFIG_SMP.
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Adrian Bunk [Fri, 4 Apr 2008 12:57:38 +0000 (14:57 +0200)]
m68k: use KBUILD_DEFCONFIG
The default defconfig should be one from arch/m68k/configs/
arch/m68k/defconfig was not exactly identical to amiga_defconfig but
also considering how long they have been without any update that doesn't
seem to have been on purpose.
Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
pata_ali: disable ATAPI DMA
libata: ATA_12/16 doesn't fall into ATAPI_MISC
libata: uninline atapi_cmd_type()
libata: fix IDENTIFY order in ata_bus_probe()
Mikulas Patocka noted that the optimization where we check if a buffer
was already dirty (and we avoid re-dirtying it) was not really SMP-safe.
Since the read of the old status was not synchronized with anything, an
aggressive CPU re-ordering of memory accesses might have moved that read
up to before the data was even written to the buffer, and another CPU
that cleaned it again, causing the newly dirty state to never actually
hit the disk.
Admittedly this would probably never trigger in practice, but it's still
wrong.
Mikulas sent a patch that fixed the problem, but I dislike the subtlety
of the whole optimization, so this is an alternate fix that is more
explicit about the particular SMP ordering for the optimization, and
separates out the speculative reads of the buffer state into its own
conditional (and makes the memory barrier only happen if we are likely
to actually hit the optimized case in the first place).
I considered removing the optimization entirely, but Andrew argued for
it's continued existence. I'm a push-over.
parport_pc: make sure to release IO ports after probing for IT87XX
Commit f63fd7e299ee13da071ecfce2b90b58c5e1562b1 ("parport_pc: detection
for SuperIO IT87XX POST") only released the IO port region on success,
not when the probe for the IT87XX chip failed.
That caused not only a reserved region to leak, but also caused an oops
when the driver module was unloaded and somebody tried to cat
/proc/ioports - because the string that was assigned to the IO port
region was a static string in the module virtual address area.
Reported-by: Lubos Lunak <l.lunak@suse.cz> Cc: Jan Kara <jack@suse.cz> Cc: Petr Cvek <petr.cvek@tul.cz> Acked-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Al Viro [Wed, 26 Mar 2008 05:57:12 +0000 (05:57 +0000)]
fix endian lossage in forcedeth
a) if you initialize something with le32_to_cpu(...), then |= it
with host-endian and feed to cpu_to_le32(), it's most definitely
*not* __le32. As sparse would've told you...
b) the whole sequence is |= cpu_to_le32(host-endian constant)
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jeff Garzik <jeff@garzik.org>
have been identified to cause a regression on some platforms due to
the assignement of legacy IRQs which makes the legacy devices
connected to those IRQs disfunctional.
Revert them.
This fixes http://bugzilla.kernel.org/show_bug.cgi?id=10382
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Thomas Gleixner [Tue, 1 Apr 2008 17:45:18 +0000 (19:45 +0200)]
x86: tsc prevent time going backwards
We already catch most of the TSC problems by sanity checks, but there
is a subtle bug which has been in the code for ever. This can cause
time jumps in the range of hours.
This was reported in:
http://lkml.org/lkml/2007/8/23/96
and
http://lkml.org/lkml/2008/3/31/23
I was able to reproduce the problem with a gettimeofday loop test on a
dual core and a quad core machine which both have sychronized
TSCs. The TSCs seems not to be perfectly in sync though, but the
kernel is not able to detect the slight delta in the sync check. Still
there exists an extremly small window where this delta can be observed
with a real big time jump. So far I was only able to reproduce this
with the vsyscall gettimeofday implementation, but in theory this
might be observable with the syscall based version as well.
CPU 0 updates the clock source variables under xtime/vyscall lock and
CPU1, where the TSC is slighty behind CPU0, is reading the time right
after the seqlock was unlocked.
The clocksource reference data was updated with the TSC from CPU0 and
the value which is read from TSC on CPU1 is less than the reference
data. This results in a huge delta value due to the unsigned
subtraction of the TSC value and the reference value. This algorithm
can not be changed due to the support of wrapping clock sources like
pm timer.
The huge delta is converted to nanoseconds and added to xtime, which
is then observable by the caller. The next gettimeofday call on CPU1
will show the correct time again as now the TSC has advanced above the
reference value.
To prevent this TSC specific wreckage we need to compare the TSC value
against the reference value and return the latter when it is larger
than the actual TSC value.
I pondered to mark the TSC unstable when the readout is smaller than
the reference value, but this would render an otherwise good and fast
clocksource unusable without a real good reason.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@elte.hu>
projects / linux-beck.git / log