During shutdown, it is impossible to reliably disable the timer and
reset_task threads. Each thread can schedule the other, which leads to
shutdown code that chases its tail.
To fix the problem, this patch removes the ability of tg3_reset_task to
schedule a new timer thread. To support this change, tg3_timer no
longer terminates itself, but rather goes into a polling mode.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is possible for multiple threads in the tg3 driver to each attempt to
schedule a run of tg3_reset_task(). The multiple tg3_reset_task
executions could all wind up on the same queue (and execute serially) or
wind up on the queues of another processor (which could execute in
parallel). Either scenario is not what was truly desired.
This patch adds a new flag, TG3_FLAG_RESET_TASK_PENDING, and uses it to
determine whether or not to schedule another run of tg3_reset_task().
With the new flag comes two new functions to facilitate scheduling and
descheduling of tg3_reset_task().
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds code to attempt to obtain the PCI function number from
the device rather than accept the number handed by the kernel. In
pass-through scenarios, the function number handed by the kernel may not
reflect the true function of the device.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On the error recovery resource unwind path, it is possible for the
driver to attempt to unmap a fragment that hadn't been mapped. This
patch fixes the problem by correcting the "last" parameter supplied.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The new 4k tx bd segmentation code had a bug in the error cleanup path.
If the driver did not map all the physical fragments, the abort path
would wind up advancing the producer index beyond the point where the
setup code stopped. This would ultimately turn into a tx recovery error
where the driver would expect the skb pointer to be set when it isn't.
This patch fixes the problem, and then makes the code a little easier to
understand.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
APE mutex register blocks are shared by all ports of multiport devices.
For some mutexing purposes, each function is assigned their own
register. For other cases, each function is assigned its own request
and grant bits of a single register. For the latter cases, the tg3
driver is incorrectly allowing each function to use the same set of
grant / request bits. This patch fixes the code so that each function
uses the appropriate bitset.
Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'for-linus' of git://github.com/cmetcalf-tilera/linux-tile:
arch/tile: factor out <arch/opcode.h> header
arch/tile: add the <arch> headers to the set of installed kernel headers
arch/tile: avoid exporting a symbol no longer used by gcc
arch/tile: avoid ISO namespace pollution with <asm/sigcontext.h>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
IB/qib: Fix panic in RC error flushing logic
IB/iser: DMA unmap TX bufs used for iSCSI/iSER headers
IB/iser: Use separate buffers for the login request/response
IB/mthca: Fix buddy->num_free allocation size
After commit 38829dc9d7 ("Documentation/CodingStyle: flesh out if-else
examples") highlight that if _only_one_ branch of a conditional
statement is a single statement, then braces are to be used on both
branches.
Signed-off-by: Antonio Ospite <ospite@studenti.unina.it>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The warning really shouldn't happen, but until we
find the reason why it does don't spew it all the
time, just once is enough to know we've hit it.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The following panic can occur when flushing a QP:
RIP: 0010:[<ffffffffa0168e8b>] [<ffffffffa0168e8b>] qib_send_complete+0x3b/0x190 [ib_qib]
RSP: 0018:ffff8803cdc6fc90 EFLAGS: 00010046
RAX: 0000000000000000 RBX: ffff8803d84ba000 RCX: 0000000000000000
RDX: 0000000000000005 RSI: ffffc90015a53430 RDI: ffff8803d84ba000
RBP: ffff8803cdc6fce0 R08: ffff8803cdc6fc90 R09: 0000000000000001
R10: 00000000ffffffff R11: 0000000000000000 R12: ffff8803d84ba0c0
R13: ffff8803d84ba5cc R14: 0000000000000800 R15: 0000000000000246
FS: 0000000000000000(0000) GS:ffff880036600000(0000) knlGS:0000000000000000
CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b
CR2: 0000000000000034 CR3: 00000003e44f9000 CR4: 00000000000406f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process qib/0 (pid: 1350, threadinfo ffff8803cdc6e000, task ffff88042728a100)
Stack:
53544c5553455201 0000000100000005 0000000000000000 ffff8803d84ba000
0000000000000000 0000000000000000 0000000000000000 0000000000000000
0000000000000000 0000000000000001 ffff8803cdc6fd30 ffffffffa0165d7a
Call Trace:
[<ffffffffa0165d7a>] qib_make_rc_req+0x36a/0xe80 [ib_qib]
[<ffffffffa0165a10>] ? qib_make_rc_req+0x0/0xe80 [ib_qib]
[<ffffffffa01698b3>] qib_do_send+0xf3/0xb60 [ib_qib]
[<ffffffff814db757>] ? thread_return+0x4e/0x777
[<ffffffffa01697c0>] ? qib_do_send+0x0/0xb60 [ib_qib]
[<ffffffff81088bf0>] worker_thread+0x170/0x2a0
[<ffffffff8108e530>] ? autoremove_wake_function+0x0/0x40
[<ffffffff81088a80>] ? worker_thread+0x0/0x2a0
[<ffffffff8108e1c6>] kthread+0x96/0xa0
[<ffffffff8100c1ca>] child_rip+0xa/0x20
[<ffffffff8108e130>] ? kthread+0x0/0xa0
[<ffffffff8100c1c0>] ? child_rip+0x0/0x20
RIP [<ffffffffa0168e8b>] qib_send_complete+0x3b/0x190 [ib_qib]
The RC error state flush logic in qib_make_rc_req() could return all
of the acked wqes and potentially have emptied the queue. It would
then unconditionally try return a flush completion via
qib_send_complete() for an invalid wqe, or worse a valid one that is
not queued. The panic results when the completion code tries to
maintain an MR reference count for a NULL MR.
This fix modifies logic to only send one completion per
qib_make_rc_req() call and changing the completion status from
IB_WC_SUCCESS to IB_WC_WR_FLUSH_ERR as the completions progress.
The outer loop will call as many times as necessary to flush the queue.
Reviewed-by: Ram Vepa <ram.vepa@qlogic.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The current driver never does DMA unmapping on these buffers. Fix that
by adding DMA unmapping to the task cleanup callback, and DMA mapping to
the task init function (drop the headers_initialized micro-optimization).
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Matt Carlson
David S. Miller
Linus Torvalds
Wang Sheng-Hui
Bryan Wu
Marcos Paulo Souza
Antonio Ospite
Kumar Appaiah
Li Zefan
John W. Linville
Johannes Berg
Roland Dreier
Mike Marciniszyn
Or Gerlitz