The iWARP connection manager uses the ib_addr services to do route
resolution (neighbour discovery in the IP world). The ib_addr
netevent callback routine, however, currently only acts on InfiniBand
neighbour updates. It needs to act on ethernet neighbour updates as
well.
This patch just removes filtering on device type altogether and will
trigger on any neighour updates where the nud_type is valid. This
simplifies the code some.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Make the untyped data region in ib_user_mad have type u64 so that it
gets aligned properly. This avoids alignment faults in ib_umad when
casting the data field to an rmpp_mad and accessing the 64-bit tid
field on architectures like ia64.
Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When there is a call to send_tsk_mgmt SRP posts a send and waits for 5
seconds to get a response.
When the QP is in the error state it is obvious that there will be no
response so it is quite useless to wait. In fact, the timeout causes
SRP to wait a long time to reconnect when a QP error occurs. (Each
abort and each reset_device calls send_tsk_mgmt, which waits for the
timeout). The following patch solves this problem by identifying the
failure and returning an immediate error code.
Signed-off-by: Ishai Rabinovitz <ishai@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
struct ib_wc currently only includes the local QP number: this matches
the IB spec, but seems mostly useless. The following patch replaces
this with the pointer to qp itself, and updates all low level drivers
and all users.
This has the following advantages:
- Ability to get a per-qp context through wc->qp->qp_context
- Existing drivers already have the qp pointer ready in poll cq, so
this change actually saves a tiny bit (extra memory read) on data path
(for ehca it would actually be expensive to find the QP pointer when
polling a CQ, but ehca does not support SRQ so we can leave wc->qp as
NULL for ehca)
- Users that need the QP number can still get it through wc->qp->qp_num
Use case:
In IPoIB connected mode code, I have a common CQ shared by multiple
QPs. To track connection usage, I need a way to get at some per-QP
context upon the completion, and I would like to avoid allocating
context object per work request just to stick a QP pointer into it.
With this code, I can just use wc->qp->qp_context.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
<rdma/ib_verbs.h> uses struct kref, so it should include <linux/kref.h>
explicitly to avoid hidden include dependencies.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Since we actively avoid highmem, calling kmap_atomic() instead
of page_address() is effectively only obfuscation.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Since we actively avoid highmem, calling kmap_atomic() instead
of page_address() is effectively only obfuscation.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Since we actively avoid highmem, calling kmap_atomic() instead
of page_address() is effectively only obfuscation.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
MMC high-speed, wide bus support and SD high-speed
are functions that aren't critical for correct
operation of the card. As such, they shouldn't mark
the card as bad or dead when there is a failure
activating these features.
This is needed in particular on some really stupid
hardware (e.g. Winbond's) where not all data transfer
commands are supported.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
The wbsd hardware is so incredibly brain damaged that it has an internal
list of commands that result in data transfers. The result being that
commands that aren't on this list aren't supported.
Instead of locking up, waiting for a data interrupt that will never come,
we try to fail a bit more gracefully.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Many controllers have an upper limit on the number of blocks that can be
transferred in one request. Allow the host drivers to specify this and make
sure we avoid hitting this limit.
Also change the max_sectors field to avoid confusion. This makes it map
less directly to the block layer limits, but as they didn't apply directly
on MMC cards anyway, this isn't a great loss.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Most controllers have an upper limit on the block size. Allow the host
drivers to specify this and make sure we avoid hitting this limit.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Fix some spaces and tabs. No semantic changes are introduced.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
This patch also adds symbolic defines for supported pci ids.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
As there's only one work item (media_switcher) to handle and it's effectively
serialized with itself, I found it more convenient to use kthread instead of
workqueue. This also allows for a working implementation of suspend/resume,
which were totally broken in the past version.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Hardware does not say whether card was inserted or removed when reporting
socket events. Moreover, during suspend, media can be removed or switched
to some other card type without notification. Therefore, for each socket
in the change set the following is performed:
1. If there's active device in the socket it's unregistered
2. Media detection is performed
3. If detection recognizes supportable media, new device is registered
This patch also alters some macros and variable names to enhance clarity.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Eject function can take advantage of the socket_id field instead of explicit
pointer comparison.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
This patch introduces no semantic changes - it is here for estetic purposes.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
In order to support correct suspend and resume several changes were needed:
1. Switch from work_struct to tasklet for command handling. When device
suspend is called workqueues are already frozen and can not be used.
2. Separate host initialization code from driver's probe and don't rely
on interrupts for host initialization. This, in turn, addresses two
problems:
a) Resume needs to re-initialize the host, but can not assume that
device interrupts were already re-armed.
b) Previously, probe will return successfully before really knowing
the state of the host, as host interrupts were not armed in time.
Now it uses polling to determine the real host state before returning.
3. Separate termination code from driver's remove. Termination may be caused
by resume, if media changed type or became unavailable during suspend.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
The register access order when setting hardware timeout was incorrect and
causing problems (wrong timeout intervals). This is now fixed.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Two changes are introduced to software timeout handler in order to simplify
its management:
1. The implementation is switched from work_struct to timer
2. Previously, software timeout was rearmed with each interrupt. Now,
current request must complete entirely within timeout interval.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Data buffer for PIO transfer used to be mapped in advance with kmap.
Abolish it in favor of on-demand kmap_atomic.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Previously, stop command was issued right after BRS (block received/sent)
event. Stop command completion event could interfere with the card busy
event, causing miscount of the written blocks.
This patch ensures that stop command issued as last action for a
particular command, after DMA sompletion event and written block
count verification.
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Thanks to the generous donation of an SDHC card by John Gilmore, and
the surprisingly enlightened decision by the SD Card Association to
publish useful specs, I've been able to bash out support for SDHC. The
changes are not too profound:
i) Add a card flag indicating the card uses block level addressing and
check it in the block driver. As we never took advantage of byte-level
addressing, this simply involves skipping the block -> byte
translation when sending commands.
ii) The layout of the CSD is changed - a set of fields are discarded
to make space for a larger C_SIZE. We did not reference any of the
discarded fields except those related to the C_SIZE.
iii) Read and write timeouts are fixed values and not calculated from
CSD values.
iv) Before invoking SEND_APP_OP_COND, we must invoke the new
SEND_IF_COND to inform the card we support SDHC.
Signed-off-by: Philipl Langdale <philipl@overt.org>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Support for these devices was broken for 2.6.18-rc1 and later by commit
146ad66eac, which added voltage level support.
This restores the previous behaviour for these devices by ensuring that when
the voltage is changed, only one write to set the voltage is performed.
It may be that both writes are needed if the voltage is being changed between
two non-zero values or that it's safe to ensure that only one write is done
if the hardware only supports one voltage; I don't know whether either is the
case nor can I test since I have only the one SD reader (1524:0550), and it
supports just the one voltage.
Signed-off-by: Darren Salt <linux@youmustbejoking.demon.co.uk>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Change the parent of cards to be a specific host (a class
device), not the physical controller. This is particularly
useful when the hardware has multiple slots, meaning
multiple hosts.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
As card_busy was only used to indicate if the host was exclusively
claimed and not really used to identify a particular card, replacing
it with just a boolean makes things a lot more easily understandable.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
When calling into the EFI firmware, the parameters need to be passed on
the stack. The recent change to use -mregparm=3 breaks x86 EFI support.
This patch is needed to allow the new Intel-based Macs to suspend to ram
(efi.get_time is called during the suspend phase).
Signed-off-by: Frederic Riss <frederic.riss@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
That code doesn't do what its author apparently thought it would do...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
[SCSI] sd: udev accessing an uninitialized scsi_disk field results in a crash
[SCSI] st: A MTIOCTOP/MTWEOF within the early warning will cause the file number to be incorrect
[SCSI] qla4xxx: bug fixes
[SCSI] Fix scsi_add_device() for async scanning
The SN Altix platform does not conform to the IOSAPIC IRQ routing model.
Add code in acpi_unregister_gsi() to check if (acpi_irq_model ==
ACPI_IRQ_MODEL_PLATFORM) and return.
Due to an oversight, this code was not added previously when
similar code was added to acpi_register_gsi().
http://marc.theaimsgroup.com/?l=linux-acpi&m=116680983430121&w=2
Signed-off-by: John Keller <jpk@sgi.com>
Acked-by: Len Brown <lenb@kernel.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Andrew Vasquez is reporting as-iosched oopses and a 65% throughput
slowdown due to the recent special-casing of direct-io against
blockdevs. We don't know why either of these things are occurring.
The patch minimally reverts us back to the 2.6.19 code for a 2.6.20
release.
Cc: Andrew Vasquez <andrew.vasquez@qlogic.com>
Cc: Ken Chen <kenchen@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We went and named them __NR_sys_foo instead of __NR_foo.
It may be too late to change this, but we can at least add the proper names
now.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
smc911x_phy_configure's error handling unconditionally unlocks the
spinlock even if it wasn't locked. Patch fixes it.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch fixes up ia64 kexec support for HP rx2620 hardware. It does
this by skipping migration of already disabled irqs. This is most likely a
problem on other ia64 platforms as well, but I've only been able to
reproduce it on one machine so far.
The full story is that handle_bad_irq() gets invoked before starting the
new kernel without this patch. This seems to happen when fixup_irqs()
calls generic_handle_irq() on already migrated (and disabled) irqs. So by
avoiding migration of disabled irqs we stay away of handle_bad_irq().
The code has been tested on three different ia64 machines, all with good
results. It is possible to trigger the same bug by offlining a processor
using echo 0 > /sys/devices/system/cpu/cpuX/online.
More detailed information is available in the following mail thread:
http://lists.osdl.org/pipermail/fastboot/2007-January/thread.html#5774
Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
Acked-by: Simon Horman <horms@verge.net.au>
Acked-by: Zou, Nanhai <nanhai.zou@intel.com>
Acked-by: Jay Lan <jlan@sgi.com>
Acked-by: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
An AIO bug was reported that sleeping function is being called in softirq
context:
BUG: warning at kernel/mutex.c:132/__mutex_lock_common()
Call Trace:
[<a000000100577b00>] __mutex_lock_slowpath+0x640/0x6c0
[<a000000100577ba0>] mutex_lock+0x20/0x40
[<a0000001000a25b0>] flush_workqueue+0xb0/0x1a0
[<a00000010018c0c0>] __put_ioctx+0xc0/0x240
[<a00000010018d470>] aio_complete+0x2f0/0x420
[<a00000010019cc80>] finished_one_bio+0x200/0x2a0
[<a00000010019d1c0>] dio_bio_complete+0x1c0/0x200
[<a00000010019d260>] dio_bio_end_aio+0x60/0x80
[<a00000010014acd0>] bio_endio+0x110/0x1c0
[<a0000001002770e0>] __end_that_request_first+0x180/0xba0
[<a000000100277b90>] end_that_request_chunk+0x30/0x60
[<a0000002073c0c70>] scsi_end_request+0x50/0x300 [scsi_mod]
[<a0000002073c1240>] scsi_io_completion+0x200/0x8a0 [scsi_mod]
[<a0000002074729b0>] sd_rw_intr+0x330/0x860 [sd_mod]
[<a0000002073b3ac0>] scsi_finish_command+0x100/0x1c0 [scsi_mod]
[<a0000002073c2910>] scsi_softirq_done+0x230/0x300 [scsi_mod]
[<a000000100277d20>] blk_done_softirq+0x160/0x1c0
[<a000000100083e00>] __do_softirq+0x200/0x240
[<a000000100083eb0>] do_softirq+0x70/0xc0
See report: http://marc.theaimsgroup.com/?l=linux-kernel&m=116599593200888&w=2
flush_workqueue() is not allowed to be called in the softirq context.
However, aio_complete() called from I/O interrupt can potentially call
put_ioctx with last ref count on ioctx and triggers bug. It is simply
incorrect to perform ioctx freeing from aio_complete.
The bug is trigger-able from a race between io_destroy() and aio_complete().
A possible scenario:
cpu0 cpu1
io_destroy aio_complete
wait_for_all_aios { __aio_put_req
... ctx->reqs_active--;
if (!ctx->reqs_active)
return;
}
...
put_ioctx(ioctx)
put_ioctx(ctx);
__put_ioctx
bam! Bug trigger!
The real problem is that the condition check of ctx->reqs_active in
wait_for_all_aios() is incorrect that access to reqs_active is not
being properly protected by spin lock.
This patch adds that protective spin lock, and at the same time removes
all duplicate ref counting for each kiocb as reqs_active is already used
as a ref count for each active ioctx. This also ensures that buggy call
to flush_workqueue() in softirq context is eliminated.
Signed-off-by: "Ken Chen" <kenchen@google.com>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Suparna Bhattacharya <suparna@in.ibm.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: <stable@kernel.org>
Acked-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix this by letting NF_CONNTRACK_H323 depend on (IPV6 || IPV6=n).
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
CC net/netfilter/nf_conntrack_netlink.o
net/netfilter/nf_conntrack_netlink.c: In function 'ctnetlink_conntrack_event':
net/netfilter/nf_conntrack_netlink.c:392: error: 'struct nf_conn' has no member named 'mark'
make[3]: *** [net/netfilter/nf_conntrack_netlink.o] Error 1
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
