- Add functionality to run in polled mode only. Includes run time
attribute to enable mode.
- Enable runtime writable hba settings for coallescing and delay parameters
Customers have requested a mode in the driver to run strictly polled.
This is generally to support an environment where the server is extremely
loaded and is looking to reclaim some cpu cycles from adapter interrupt
handling.
This patch adds a new "poll" attribute, and the following behavior:
if value is 0 (default):
The driver uses the normal method for i/o completion. It uses the
firmware feature of interrupt coalesing. The firmware allows a
minimum number of i/o completions before an interrupt, or a maximum
time delay between interrupts. By default, the driver sets these
to no delay (disabled) or 1 i/o - meaning coalescing is disabled.
Attributes were provided to change the coalescing values, but it was
a module-load time only and global across all adapters.
This patch allows them to be writable on a per-adapter basis.
if value is 1 :
Interrupts are left enabled, expecting that the user has tuned the
interrupt coalescing values. When this setting is enabled, the driver
will attempt to service completed i/o whenever new i/o is submitted
to the adapter. If the coalescing values are large, and the i/o
generation rate steady, an interrupt will be avoided by servicing
completed i/o prior to the coalescing thresholds kicking in. However,
if the i/o completion load is high enough or i/o generation slow, the
coalescion values will ensure that completed i/o is serviced in a timely
fashion.
if value is 3 :
Turns off FCP i/o interrupts altogether. The coalescing values now have
no effect. A new attribute "poll_tmo" (default 10ms) exists to set
the polling interval for i/o completion. When this setting is enabled,
the driver will attempt to service completed i/o and restart the
interval timer whenever new i/o is submitted. This behavior allows for
servicing of completed i/o sooner than the interval timer, but ensures
that if no i/o is being issued, then the interval timer will kick in
to service the outstanding i/o.
Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- Release task management command before counting outstanding commands.
TMF was being erroneously counted as an active outstanding command.
- Serialize EH calls and block requests when EH function is running.
Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
[SCSI] lpfc 8.1.1 : Adjust use of scsi_block_requests and interaction w/ FC transport
- Remove unnecessary scsi_block_requests calls on rport deletes.
This was deadlocking the sdev removals as they wanted to flush commands.
- No longer block requests when adding the remote port (to block
discovery). Instead, register, then change port role. Maps to Qlogic
behavior, and closer to the register-node-upon-first-ELS behavior.
Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Cause: Link bounces were causing discovery ELS's to be killed.
Driver was not properly flushing ELS commands upon the subsequent
link bounces. Thus, processing of ELS post link bounce erroneously
assumed discovery failure and device loss.
Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Miscellaneous Cleanups:
- Remove ProgType READ_REV mailbox command value check in lpfc_config_port_prep.
- Convert simple printk to an lpfc_printf_log in queuecommand.
- Modify lpfc_abort_handler message 0749 to display more accurate text and data.
- Minor style cleanup: fix 3 long lines in lpfc_hw.h
Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Mike Christie [Wed, 30 Nov 2005 08:27:19 +0000 (02:27 -0600)]
[SCSI] iscsi: check header digests for mgmt tasks
From Wang Zhenyu:
check header digest for cmd and mgmt tasks
Signed-off-by: Wang Zhenyu <zhenyu.z.wang@intel.com> Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: Alex Aizman <itn780@yahoo.com> Signed-off-by: Dmitry Yusupov <dmitry_yus@yahoo.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Mike Christie [Wed, 30 Nov 2005 05:12:49 +0000 (23:12 -0600)]
[SCSI] iscsi: redirect fix
From tomof@acm.org:
There is one more issue about Equallogic systems. They send
re-direction info with FIN. I think that the kernel module needs to
let iscsid to read data from the socket before killing it.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: Alex Aizman <itn780@yahoo.com> Signed-off-by: Dmitry Yusupov <dmitry_yus@yahoo.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Matthew Wilcox [Wed, 30 Nov 2005 04:08:44 +0000 (23:08 -0500)]
[SCSI] sym2: Allow NVRAM settings to limit speed and width
The NVRAM for both Tekram and Symbios boards allows the user to set the
speed and width for individual targets. I took that code out in March
2004 when we introduced Domain Validation, but it seems there's still
a legitimate need for it in some configurations.
Signed-off-by: Matthew Wilcox <matthew@wil.cx> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Matthew Wilcox [Wed, 30 Nov 2005 04:08:33 +0000 (23:08 -0500)]
[SCSI] sym2: Remove code to handle DMA_BIDIRECTION requests
The upper layer doesn't send these down since 2.4.x (or 2.6 in
practice), so no need to handle it. Inline sym_setup_data_pointers
into its only caller so we can fail gracefully in the case we'd get
one neverless.
Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Matthew Wilcox <matthew@wil.cx> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Andrew Vasquez [Wed, 9 Nov 2005 23:49:04 +0000 (15:49 -0800)]
[SCSI] qla2xxx: Add full firmware(-request) hotplug support for all ISPs.
Transition driver to exclusively use the request_firmware()
interfaces to retrieve firmware-blobs from user-space. This
will be the default behaviour going forward until the
embedded firmware-binary images are removed from the
upstream kernel.
Upon request, the driver caches the firmware image until the
driver is unloaded.
NOTE: The option is present to allow the user to continue to
use the firmware-loader modules, but, should be considered
deprecated.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Rejections fixed up and Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Arjan van de Ven [Mon, 28 Nov 2005 15:22:25 +0000 (16:22 +0100)]
[SCSI] Mark some core scsi data structures const
patch below marks a few scsi core datastructures as const, so that they end up
in the .rodata section and don't cacheline share with things that get dirtied
Signed-off-by: Arjan van de Ven <arjan@infradead.org> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
[SCSI] ipr: Driver initialization fix for kexec/kdump
When kexec booting a kernel when the previous kernel did not
call ipr's shutdown method, the ipr adapter does not get
properly initialized, which can result in the ipr adapter
completing commands issued by the previous kernel. Fix ipr
to detect this scenario by reading the adapter's interrupt
mask register and the microprocessor interrupt register.
If the interrupt mask register indicates that interrupts
are enabled or the reset alert bit is set when the card is
probed, this means the card is in an unknown state and we
hard reset the card.
Signed-off-by: Brian King <brking@us.ibm.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
[PATCH] fbdev: Fix incorrect unaligned access in little-endian machines
The drawing function cfbfillrect does not work correctly when access is not
unsigned-long aligned. It manifests as extra lines of pixels that are not
complete drawn. Reversing the shift operator solves the problem, so I would
presume that this bug would manifest only on little endian machines. The
function cfbcopyarea may also have this bug.
[PATCH] fbdev: Shift pixel value before entering loop in cfbimageblit
In slow imageblit, the pixel value is shifted by a certain amount (dependent
on the bpp and endianness) for each iteration. This is inefficient. Better
do the shifting once before going into the loop.
Knut Petersen [Tue, 13 Dec 2005 06:17:19 +0000 (22:17 -0800)]
[PATCH] fbdev: fix switch to KD_TEXT, enhanced version
Every framebuffer driver relies on the assumption that the set_par()
function of the driver is called before drawing functions and other
functions dependent on the hardware state are executed.
Whenever you switch from X to a framebuffer console for the very first
time, there is a chance that a broken X system has _not_ set the mode to
KD_GRAPHICS, thus the vt and framebuffer code executes a screen redraw and
several other functions before a set_par() is executed. This is believed
to be not a bug of linux but a bug of X/xdm. At least some X releases used
by SuSE and Debian show this behaviour.
There was a 2nd case, but that has been fixed by Antonino Daplas on
10-dec-2005.
This patch allows drivers to set a flag to inform fbcon_switch() that they
prefer a set_par() call on every console switch, working around the
problems caused by the broken X releases.
The flag will be used by the next release of cyblafb and might help other
drivers that assume a hardware state different to the one used by X.
As the default behaviour does not change, this patch should be acceptable
to everybody.
Signed-off-by: Knut Petersen <Knut_Petersen@t-online.de> Acked-by: "Antonino A. Daplas" <adaplas@pol.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] fbcon: Add ability to save/restore graphics state
Add hooks to save and restore the graphics state. These hooks are called in
fbcon_blank() when entering/leaving KD_GRAPHICS mode. This is needed by
savagefb at least so it can cooperate with savage_dri and by cyblafb.
[PATCH] fbcon: fix complement_mask() with 512 character map
There is a bug in the complement_mask when you have a 512-character map.
Linux boots to a default 256-character map and most probably your login
profile is loading a 512-character map which results in a bad gpm cursor.
Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Mike Miller [Tue, 13 Dec 2005 06:17:14 +0000 (22:17 -0800)]
[PATCH] cciss: fix for deregister_disk
This patch adds setting our drv->queue = NULL back in deregister_disk. The
drv->queue is part of our controller struct. blk_cleanup_queue works only
on the queue in the gendisk struct.
Signed-off-by: Mike Miller <mike.miller@hp.com> Cc: Jens Axboe <axboe@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Eric Dumazet [Tue, 13 Dec 2005 06:17:14 +0000 (22:17 -0800)]
[PATCH] x86_64: Bug correction in populate_memnodemap()
As reported by Keith Mannthey, there are problems in populate_memnodemap()
The bug was that the compute_hash_shift() was returning 31, with incorrect
initialization of memnodemap[]
To correct the bug, we must use (1UL << shift) instead of (1 << shift) to
avoid an integer overflow, and we must check that shift < 64 to avoid an
infinite loop.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
john stultz [Tue, 13 Dec 2005 06:17:13 +0000 (22:17 -0800)]
[PATCH] x86_64: Fix collision between pmtimer and pit/hpet
On systems that do not support the HPET legacy functions (basically the IBM
x460, but there could be others), in time_init() we accidentally fall into a
PM timer conditional and set the vxtime_hz value to the PM timer's frequency.
We then use this value with the HPET for timekeeping.
This patch (which mimics the behavior in time_init_gtod) corrects the
collision.
Andi Kleen [Tue, 13 Dec 2005 06:17:11 +0000 (22:17 -0800)]
[PATCH] i386/x86-64 Correct for broken MCFG tables on K8 systems
They report all busses as MMCONFIG capable, but it never works for the
internal devices in the CPU's builtin northbridge.
It just probes all func 0 devices on bus 0 (the internal northbridge is
currently always on bus 0) and if they are not accessible using MCFG they are
put into a special fallback bitmap.
On systems where it isn't we assume the BIOS vendor supplied correct MCFG.
Requires the earlier patch for mmconfig type1 fallback
Andi Kleen [Tue, 13 Dec 2005 06:17:10 +0000 (22:17 -0800)]
[PATCH] i386/x86-64 Fall back to type 1 access when no entry found
When there is no entry for a bus in MCFG fall back to type1. This is
especially important on K8 systems where always some devices can't be accessed
using mmconfig (in particular the builtin northbridge doesn't support it for
its own devices)
Andi Kleen [Tue, 13 Dec 2005 06:17:09 +0000 (22:17 -0800)]
[PATCH] i386/x86-64: Don't call change_page_attr with a spinlock held
It's illegal because it can sleep.
Use a two step lookup scheme instead. First look up the vm_struct, then
change the direct mapping, then finally unmap it. That's ok because nobody
can change the particular virtual address range as long as the vm_struct is
still in the global list.
Also added some LinuxDoc documentation to iounmap.
Shaohua Li [Tue, 13 Dec 2005 06:17:08 +0000 (22:17 -0800)]
[PATCH] i386/x86-64 disable LAPIC completely for offline CPU
Disabling LAPIC timer isn't sufficient. In some situations, such as we
enabled NMI watchdog, there is still unexpected interrupt (such as NMI)
invoked in offline CPU. This also avoids offline CPU receives spurious
interrupt and anything similar.
Signed-off-by: Shaohua Li <shaohua.li@intel.com> Signed-off-by: Andi Kleen <ak@suse.de> Acked-by: "Seth, Rohit" <rohit.seth@intel.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Dave Airlie [Tue, 13 Dec 2005 04:18:41 +0000 (04:18 +0000)]
[drm] fix radeon aperture issue
Ben noticed that on certain cards we've landed the AGP space on top of
the second aperture instead of after it.. Which messes things up a lot
on those machines.
This just moves the gart further out, a more correct fix is in the works
from Ben for after 2.6.15.
Signed-off-by: Dave Airlie <airlied@linux.ie> CC: Ben Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Get rid of warning in case of race with ring full and lockless
tx on the skge driver. It is possible to be in the transmit
routine with no available slots and already stopped.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org> Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Mark Lord [Tue, 13 Dec 2005 04:19:28 +0000 (23:19 -0500)]
[PATCH] libata-core.c: fix parameter bug on kunmap_atomic() calls
Fix incorrect pointer usage on two calls to kunmap_atomic().
This seems to happen a lot, because kunmap() wants the struct page *,
whereas kunmap_atomic() instead wants the mapped virtual address.
Signed-off-by: Mark Lord <liml@rtr.ca> Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Linus Torvalds [Tue, 13 Dec 2005 00:24:33 +0000 (16:24 -0800)]
get_user_pages: don't try to follow PFNMAP pages
Nick Piggin points out that a few drivers play games with VM_IO (why?
who knows..) and thus a pfn-remapped area may not have that bit set even
if remap_pfn_range() set it originally.
So make it explicit in get_user_pages() that we don't follow VM_PFNMAP
pages, since pretty much by definition they do not have a "struct page"
associated with them.
Marcus Sundberg [Mon, 12 Dec 2005 23:02:48 +0000 (15:02 -0800)]
[NETFILTER]: ip_nat_tftp: Fix expectation NAT
When a TFTP client is SNATed so that the port is also changed, the
port is never changed back for the expected connection.
Signed-off-by: Marcus Sundberg <marcus@ingate.com> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Paul Jackson [Mon, 12 Dec 2005 22:42:31 +0000 (14:42 -0800)]
[SPARC]: atomic_clear_mask build fix
This fixes one build error introduced in sparc with the patch of Oct 30,
resent Nov 4 "[patch 3/5] atomic: atomic_inc_not_zero" I still can't get
sparc to build, but at least it gets further after I remove this line.
Apparently, this change was agreed to by Andrew and Nick on Nov 14, but
everyone thought someone else was doing it.
Signed-off-by: Paul Jackson <pj@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Brian King [Mon, 12 Dec 2005 19:05:08 +0000 (13:05 -0600)]
[PATCH] Fix SCSI scanning slab corruption
There is a double free in the scsi scan code if a LLDD's slave_alloc()
call fails. There is a direct call to scsi_free_queue and then the
following put_device calls the release function, which also frees the
queue.
Remove the redundant scsi_free_queue.
Signed-off-by: Brian King <brking@us.ibm.com> Tested-by: Nathan Lynch <ntl@pobox.com>
[ Also removed some strange whitespace artifacts in that area ] Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Olaf Hering [Fri, 9 Dec 2005 18:12:10 +0000 (19:12 +0100)]
[PATCH] pcnet32: use MAC address from prom also on powerpc
The CSR contains garbage after a coldboot on RS/6000.
One some systems (like my 44p 270) the MAC address is all FF,
on others (like my B50) it is ff:ff:ff:fd:ff:6b.
It can eventually be fixed by loading pcnet32, set the interface
into the UP state, rmmod pcnet32 and load it again. But this worked
only on the 270.
Only netbooting after a cold start provides the correct MAC address
via prom and CSR. This makes it very unreliable.
I dont know why the MAC is stored in two different places. Remove
the special case for powerpc, which was added in early 2.4 development.
Signed-off-by: Olaf Hering <olh@suse.de>
drivers/net/pcnet32.c | 5 -----
1 files changed, 5 deletions(-) Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
"All it's doing is deferring the device_put() from the
scsi_put_command() to after the scsi_run_queue(), which doesn't fix
the sleep while atomic problem of the device release method. In both
cases we still get the semaphore in atomic context problem which is
caused by scsi_reap_target() doing a device_del(), which I assumed
(wrongly) was valid from atomic context."
NeilBrown [Mon, 12 Dec 2005 10:39:17 +0000 (02:39 -0800)]
[PATCH] md: use correct size of raid5 stripe cache when measuring how full it is
The raid5 stripe cache was recently changed from fixed size (NR_STRIPES) to
variable size (conf->max_nr_stripes). However there are two places that still
use the constant and as a result, reducing the size of the stripe cache can
result in a deadlock.
Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Dave Jones [Mon, 12 Dec 2005 08:37:40 +0000 (00:37 -0800)]
[PATCH] ACPI: fix sleeping whilst atomic warnings on resume
This has been broken for months. On resume, we call acpi_pci_link_set()
with interrupts off, so we get a warning when we try to do a kmalloc of non
atomic memory. The actual allocation is just 2 long's (plus extra byte for
some reason I can't fathom), so a simple conversion to GFP_ATOMIC is
probably the safest way to fix this.
The error looks like this..
Debug: sleeping function called from invalid context at mm/slab.c:2486
in_atomic():0, irqs_disabled():1
[<c0143f6c>] kmem_cache_alloc+0x40/0x56
[<c0206a2e>] acpi_pci_link_set+0x3f/0x17f
[<c0206f96>] irqrouter_resume+0x1e/0x3c
[<c0239bca>] __sysdev_resume+0x11/0x6b
[<c0239e88>] sysdev_resume+0x34/0x52
[<c023de21>] device_power_up+0x5/0xa
Signed-off-by: Dave Jones <davej@redhat.com> Cc: "Brown, Len" <len.brown@intel.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Haren Myneni [Mon, 12 Dec 2005 08:37:39 +0000 (00:37 -0800)]
[PATCH] fix in __alloc_bootmem_core() when there is no free page in first node's memory
Hitting BUG_ON() in __alloc_bootmem_core() when there is no free page
available in the first node's memory. For the case of kdump on PPC64
(Power 4 machine), the captured kernel is used two memory regions - memory
for TCE tables (tce-base and tce-size at top of RAM and reserved) and
captured kernel memory region (crashk_base and crashk_size). Since we
reserve the memory for the first node, we should be returning from
__alloc_bootmem_core() to search for the next node (pg_dat).
Currently, find_next_zero_bit() is returning the n^th bit (eidx) when there
is no free page. Then, test_bit() is failed since we set 0xff only for the
actual size initially (init_bootmem_core()) even though rounded up to one
page for bdata->node_bootmem_map. We are hitting the BUG_ON after failing
to enter second "for" loop.
Signed-off-by: Haren Myneni <haren@us.ibm.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Nicolas Pitre [Mon, 12 Dec 2005 08:37:36 +0000 (00:37 -0800)]
[PATCH] input: fix ucb1x00-ts breakage after conversion to dynamic input_dev allocation
The bd622663192e8ebebb27dc1d9397f352a82d2495 commit broke the UCB1x00
touchscreen driver since the idev structure was assumed to be into the ts
structure, simply casting the former to the later in a couple places.
This patch fixes those, and also cache the idev pointer between multiple
calls to input_report_abs() to avoid growing the compiled code needlessly.
Signed-off-by: Nicolas Pitre <nico@cam.org> Cc: Dmitry Torokhov <dtor_core@ameritech.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] kprobes: increment kprobe missed count for multiprobes
When multiple probes are registered at the same address and if due to some
recursion (probe getting triggered within a probe handler), we skip calling
pre_handlers and just increment nmissed field.
The below patch make sure it walks the list for multiple probes case.
Without the below patch we get incorrect results of nmissed count for
multiple probe case.
Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
For Kprobes critical path is the path from debug break exception handler
till the control reaches kprobes exception code. No probes can be
supported in this path as we will end up in recursion.
This patch prevents this by moving the below function to safe __kprobes
section onto which no probes can be inserted.
Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Matt Domsch [Mon, 12 Dec 2005 08:37:32 +0000 (00:37 -0800)]
[PATCH] ipmi: fix panic generator ID
The IPMI specifcation says the generator ID is 0x20, but that is for bits
7-1. Bit 0 is set to specify it is a software event. The correct value is
0x41. Without this fix, panic events written into the System Event Log
appear to come from an "unknown" generator, rather than from the kernel.
Signed-off-by: Jordan Hargrave <Jordan_Hargrave@dell.com> Signed-off-by: Matt Domsch <Matt_Domsch@dell.com> Acked-by: Corey Minyard <minyard@acm.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
projects / linux-beck.git / log