Calling zfcp_erp_strategy_check_action() after zfcp_erp_action_to_running()
in zfcp_erp_strategy() might cause an unbalanced up() for erp_ready_sem,
which makes the zfcp recovery fail somewhere along the way:
erp thread processing erp_action:
|
| someone waking up erp thread for erp_action
| |
| | someone else dismissing erp_action:
| | |
V V V
write_lock_irqsave(&adapter->erp_lock, flags);
...
if (zfcp_erp_action_exists(erp_action) == ZFCP_ERP_ACTION_RUNNING) {
zfcp_erp_action_to_ready(erp_action);
up(&adapter->erp_ready_sem); /* first up() for erp_action */
}
write_unlock_irqrestore(&adapter->erp_lock, flags);
write_lock_irqsave(&adapter->erp_lock, flags);
...
zfcp_erp_action_to_running(erp_action);
write_unlock_restore(&adapter->erp_lock, flags);
/* processing erp_action */
write_lock_irqsave(&adapter->erp_lock, flags);
...
erp_action->status |= ZFCP_STATUS_ERP_DISMISSED;
if (zfcp_erp_action_exists(erp_action) ==
ZFCP_ERP_ACTION_RUNNING) {
zfcp_erp_action_to_ready(erp_action);
up(&adapter->erp_ready_sem);
/* second, unbalanced up() for erp_action */
}
...
write_unlock_restore(&adapter->erp_lock, flags);
write_lock_irqsave(&adapter->erp_lock, flags);
if (erp_action->status & ZFCP_STATUS_ERP_DISMISSED) {
zfcp_erp_action_dequeue(erp_action);
retval = ZFCP_ERP_DISMISSED;
}
...
write_unlock_restore(&adapter->erp_lock, flags);
down(&adapter->erp_ready_sem);
/* this down() is meant to balance the first up() */
The erp thread must not dismiss an erp_action after moving that action to
erp_running_head. Instead it should just go through the down() operation,
which balances the first up(), and run through zfcp_erp_strategy one more
time for the second up(), which eventually cleans up erp_action. Which
is similar to the normal processing of an event for erp_action doing
something asynchronously (e.g. waiting for the completion of an fsf_req).
This only works if we make sure that a dismissed erp_action is passed to
zfcp_erp_strategy() prior to the other action, which caused actions to be
dismissed. Therefore the patch implements this rule: running actions go to
the head of the ready list; new actions go to the tail of the ready list;
the erp thread picks actions to be processed from the ready list's head.
Signed-off-by: Martin Peschke <mp3@de.ibm.com>
Acked-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
zfcp_erp_action_dismiss() used to ignore any actions in the ready list. This
is a bug. Any action superseded by a stronger action needs to be dismissed.
This patch changes zfcp_erp_action_dismiss() so that it dismisses actions
regardless of their list affiliation. The ERP thread is able to handle this.
It is important to kick the erp thread only for actions in the running list,
though, as an imbalance of wakeup signals would confuse the erp thread
otherwise.
Signed-off-by: Martin Peschke <mp3@de.ibm.com>
Acked-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Most drivers need to set length and offset as well, so may as well fold
those three lines into one.
Add sg_assign_page() for those two locations that only needed to set
the page, where the offset/length is set outside of the function context.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Cleanup the whitepace from the entire zfcp driver to prevent
to have those changes in future feature or function patches.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Already register the debug feature before the zfcp adapter is
set online. This allows to use the debug feature to investigate
the online/offline sequence.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
cleanup, using ERP request mempool for all ERP versions of
the exchange functions (exchange_config (ECD), exchange_port (EPD) )
providing individual versions of the ECD, EPD functions for ERP
and other purposes (_sync).
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Remove braces for only one statement
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
zfcp_adapter_enqueue initialized adapter->ccw_device twice with
the same value. Remove the second assignment, since it is not
necessary.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Avoid "if (whatever) ;" constructs since they seem to confuse people,
even if there is a comment.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Also removes a bunch of ^L in drivers/s390/cio/cmf.c
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Introduce a shutdown method for the ccw bus that calls the driver
specific shutdown method in struct ccw_driver.
Switch zfcp to the new ccw_driver shutdown method.
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix the data buffer accessor patch.
For request without a data buffer nothing was written into
a SBALE.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- remove the unnecessary map_single path.
- convert to use the new accessors for the sg lists and the
parameters.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Slab destructors were no longer supported after Christoph's
c59def9f22 change. They've been
BUGs for both slab and slub, and slob never supported them
either.
This rips out support for the dtor pointer from kmem_cache_create()
completely and fixes up every single callsite in the kernel (there were
about 224, not including the slab allocator definitions themselves,
or the documentation references).
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Get rid of two 'warning: Using plain integer as NULL pointer'.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Memory allocated with kmalloc is always initialzed to 0 with memset.
Replace the two calls with kzalloc, that already does both steps.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
When reporting SCSI devices to the SCSI midlayer, use the FCP LUN as
LUN reported to the SCSI layer. With this approach, zfcp does not have
to create unique LUNS, and this code can be removed.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
In some cases we did not reset some adapter status flags properly.
This patch clears these flags during FCP adapter shutdown.
Signed-off-by: Volker Sameske <sameske@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
IO stall after deleting and path checker changes after reenabling zfcp device
Setting one zfcp device offline using chccwdev in a multipath
environment and waiting will lead to IO stall on all paths.
After setting the zfcp device back online using chccwdev,
the devices with io stall will have a different path checker.
Devices corresponding to the deleted units are never freed.
This has the effect that 'slave_destroy' is never called and zfcp
still thinks that this unit is registered
(ZFCP_STATUS_UNIT_REGISTERED is still set). Hence the erp
routine is not called correctly and the unit is not enabled properly.
Do not delete rport and the sdev. Just set the host to block on
'offline'. Setting host online again will then remove the blocked status
and everything is fine again.
Signed-off-by: Michael Loehr <mloehr2@linux.vnet.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
avoid clutter in erp_dbf
cleanup zfcp_fsf_req_dismiss functions:
- avoid clutter in erp_dbf (reqs_active is always 0)
- fold called three-line function into calling function
- add meaningful comment
- coding style
Signed-off-by: Martin Peschke <mp3@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The boxed flag for units was never cleared. This doesn't hurt, but on
ACL updates the error recovery could reopen more units than needed.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Must clear adapter failed flag if an fsf request times out. This is
necessary because on link down situations the failed flags gets set
but the QDIO queues are still up. Since an adapter reopen will be
skipped if the failed flag is set an adapter_reopen that is issued
on fsf request timeout has no effect if the local link is down.
Might lead to locked up system if the SCSI stack is waiting for abort
completion.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Simplify request ID management and make sure that frequently used
functions are inlined. Also fix a memory leak in zfcp_adapter_enqueue()
which only gets hit in error handling.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The SCSI stack requires low level drivers to register and
unregister devices. For zfcp this leads to the situation where
zfcp calls the SCSI stack, the SCSI tries to scan the new device
and the scan SCSI command fails. This would require the zfcp erp,
but the erp thread is already blocked in the register call.
The fix is to make sure that the calls from the ERP thread to
the SCSI stack do not block the ERP thread. In detail:
1) Use a workqueue to avoid blocking of the scsi_scan_target calls.
2) When removing a unit make sure that no scsi_scan_target call is
pending.
3) Replace scsi_flush_work with scsi_target_unblock. This avoids
blocking and has the same result.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
There is a possible race condition while generating the unique
request ids and sequence numbers. Both might be read at the
same time and have the same value. Fix this by serializing the
access through the queue lock of the adapter: First call
zfcp_fsf_req_sbal_get that acquires the lock, then read and
increment the unique ids.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
S_ID and D_ID are defined in the FCP spec as 3 byte fields.
Change the output in zfcp print statements accordingly to print
them with only 3 bytes.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
For each request that is sent to the FCP adapter, zfcp allocates
memory. Status information and data that is being read from the
device is written to this memory by the hardware. After that,
the hardware signals this via the response queue and zfcp
continues processing.
Now, if zfcp detects that there is a signal for an incoming
response from the hardware, but there is no outstanding request
for that request id, then some memory that can be in use anywhere
in the system has just been overwritten. This should never happen,
but if it does, stop the system with a panic.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Many struct file_operations in the kernel can be "const". Marking them const
moves these to the .rodata section, which avoids false sharing with potential
dirty data. In addition it'll catch accidental writes at compile time to
these shared resources.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Invalid locking order. Kernel hangs after trying to take two locks
which are dependend on each other. Introducing temporary variable
to free requests. Free lock after requests are copied.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Run this:
#!/bin/sh
for f in $(grep -Erl "\([^\)]*\) *k[cmz]alloc" *) ; do
echo "De-casting $f..."
perl -pi -e "s/ ?= ?\([^\)]*\) *(k[cmz]alloc) *\(/ = \1\(/" $f
done
And then go through and reinstate those cases where code is casting pointers
to non-pointers.
And then drop a few hunks which conflicted with outstanding work.
Cc: Russell King <rmk@arm.linux.org.uk>, Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Greg KH <greg@kroah.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Paul Fulghum <paulkf@microgate.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Karsten Keil <kkeil@suse.de>
Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: Ian Kent <raven@themaw.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Neil Brown <neilb@cse.unsw.edu.au>
Cc: Jaroslav Kysela <perex@suse.cz>
Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
