SCSI-ml manages the queueing limits for the device and host, but
does not do so at the target level. However something something similar
can come in userful when a driver is transitioning a transport object to
the the blocked state, becuase at that time we do not want to queue
io and we do not want the queuecommand to be called again.
The patch adds code similar to the exisiting SCSI_ML_*BUSY handlers.
You can now return SCSI_MLQUEUE_TARGET_BUSY when we hit
a transport level queueing issue like the hw cannot allocate some
resource at the iscsi session/connection level, or the target has temporarily
closed or shrunk the queueing window, or if we are transitioning
to the blocked state.
bnx2i, when they rework their firmware according to netdev
developers requests, will also need to be able to limit queueing at this
level. bnx2i will hook into libiscsi, but will allocate a scsi host per
netdevice/hba, so unlike pure software iscsi/iser which is allocating
a host per session, it cannot set the scsi_host->can_queue and return
SCSI_MLQUEUE_HOST_BUSY to reflect queueing limits on the transport.
The iscsi class/driver can also set a scsi_target->can_queue value which
reflects the max commands the driver/class can support. For iscsi this
reflects the number of commands we can support for each session due to
session/connection hw limits, driver limits, and to also reflect the
session/targets's queueing window.
Changes:
v1 - initial patch.
v2 - Fix scsi_run_queue handling of multiple blocked targets.
Previously we would break from the main loop if a device was added back on
the starved list. We now run over the list and check if any target is
blocked.
v3 - Rediff for scsi-misc.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Brian King <brking@linux.vnet.ibm.com> reported that fibre channel
devices can oops during scanning if their ports block (because the
device goes from CREATED -> BLOCK -> RUNNING rather than CREATED ->
BLOCK -> CREATED).
Fix this by adding a new state: CREATED_BLOCK which can only transition
back to CREATED and disallow the CREATED -> BLOCK transition. Now both
the created and blocked states that the mid-layer recognises can include
CREATED_BLOCK.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The created and blocked states are very shortly going to correspond to
mixed sdev_state states.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
For IBM z series certain LUNs can no longer be accessed.
This is because kernel version 2.6.19 a check was introduced not to
create a generic SCSI device for devices that return PQ=1 and
PDT=0x1f. For WLUNs (see SAM-3, p. 41ff) generic SCSI devices should
be created unconditionally without looking at the PQ bit, so add a
check for WLUNs in with this test.
Acked-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
[jejb: fixed up a ton of missed conversions.
All of you are on notice this has happened, driver trees will now
need to be rebased]
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Cc: SCSI List <linux-scsi@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The mutex is released on a successful return, so it would seem that it
should be released on an error return as well.
The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@@
expression l;
@@
mutex_lock(l);
... when != mutex_unlock(l)
when any
when strict
(
if (...) { ... when != mutex_unlock(l)
+ mutex_unlock(l);
return ...;
}
|
mutex_unlock(l);
)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Make the needlessly global struct scsi_{host,target}_type static.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Some drivers have duplicated unlikely() macros. IS_ERR() already has
unlikely() in itself.
This patch cleans up such pointless code.
Signed-off-by: Hirofumi Nakagawa <hnakagawa@miraclelinux.com>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Jeff Garzik <jeff@garzik.org>
Cc: Paul Clements <paul.clements@steeleye.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Michael Halcrow <mhalcrow@us.ibm.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Carsten Otte <cotte@de.ibm.com>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Jaroslav Kysela <perex@perex.cz>
Cc: Takashi Iwai <tiwai@suse.de>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current target allocation code registeres each possible target
with sysfs; it will be deleted again if no useable LUN on this target
was found. This results in a string of 'target add/target remove' uevents.
Based on a patch by Hannes Reinecke <hare@suse.de> this patch reworks
the target allocation code so that only uevents for existing targets
are sent. The sysfs registration is split off from the existing
scsi_target_alloc() into a in a new scsi_add_target() function, which
should be called whenever an existing target is found. Only then a
uevent is sent, so we'll be generating events for existing targets
only.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This patch implements scsi_host and scsi_target device types
and adds both to the scsi_bus.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Fix docbook problem in SCSI source files.
These cause the generated docbook to be incorrect.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
While trying to convert the SPI transport class to attribute groups, I
discovered that we don't actually have any transport configure points
for either the target or the host. This patch adds these missing
transport class triggers. The host one is simply done after the add,
the target one tries to be more clever and add it after devices may have
been placed on the target (so the device configure will have set up the
target parameters).
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Some SCSI tape medium changers that need the BLIST_SINGLELUN flag have
the medium changer at one LUN and the tape drive at a different LUN.
The inquiry string of the tape drive may be different from that of the
medium changer. In order for single_lun to be effective, every
scsi_device under a given scsi_target must have it set. This means that
there needs to be a blacklist entry for BOTH the medium changer AND the
tape drive, which is impractical because some medium changers may be
paired with a variety of different tape drive models. It makes more
sense to put the single_lun flag in scsi_target instead of scsi_device,
which causes every device at a given target ID to inherit the single_lun
flag from one LUN. This makes it possible to blacklist just the medium
changer and not the tape drive.
Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Add Documentation/DocBook/scsi_midlayer.tmpl, add to Makefile, and update
lots of kerneldoc comments in drivers/scsi/*.
Updated with comments from Stefan Richter, Stephen M. Cameron,
James Bottomley and Randy Dunlap.
Signed-off-by: Rob Landley <rob@landley.net>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
If kthread_run failed, we would fail to scan the host, and leak the
allocated async_scan_data. Since using a separate thread is just an
optimisation, do the scan synchronously if we fail to spawn a thread.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
In scsi module I've found some inconsistency between variable type
used in module_param_named and type passed to module_param_named as an
argument. Especially the inconsistency of `max_scsi_luns' parameter is
a bit serious because the description text says "last scsi LUN (should
be between 1 and 2^32-1)".
Signed-off-by: Masatake YAMATO <jet@gyve.org>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Stress-testing and some thought has revealed some places where
asynchronous scanning needs some more attention to locking.
- Since async_scan is a bit, we need to hold the host_lock while
modifying it to prevent races against other CPUs modifying the word
that bit is in. This is probably a theoretical race for the moment,
but other patches may change that.
- The async_scan bit means not only that this host is being scanned
asynchronously, but that all the devices attached to this host are not
yet added to sysfs. So we must ensure that this bit is always in sync.
I've chosen to do this with the scan_mutex since it's already acquired
in most of the right places.
- If the host changes state to deleted while we're in the middle of
a scan, we'll end up with some devices on the host's list which must
be deleted. Add a check to scsi_sysfs_add_devices() to ensure the
host is still running.
- To avoid the async_scan bit being protected by three locks, the
async_scan_lock now only protects the scanning_list.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (166 commits)
[SCSI] ibmvscsi: convert to use the data buffer accessors
[SCSI] dc395x: convert to use the data buffer accessors
[SCSI] ncr53c8xx: convert to use the data buffer accessors
[SCSI] sym53c8xx: convert to use the data buffer accessors
[SCSI] ppa: coding police and printk levels
[SCSI] aic7xxx_old: remove redundant GFP_ATOMIC from kmalloc
[SCSI] i2o: remove redundant GFP_ATOMIC from kmalloc from device.c
[SCSI] remove the dead CYBERSTORMIII_SCSI option
[SCSI] don't build scsi_dma_{map,unmap} for !HAS_DMA
[SCSI] Clean up scsi_add_lun a bit
[SCSI] 53c700: Remove printk, which triggers because of low scsi clock on SNI RMs
[SCSI] sni_53c710: Cleanup
[SCSI] qla4xxx: Fix underrun/overrun conditions
[SCSI] megaraid_mbox: use mutex instead of semaphore
[SCSI] aacraid: add 51245, 51645 and 52245 adapters to documentation.
[SCSI] qla2xxx: update version to 8.02.00-k1.
[SCSI] qla2xxx: add support for NPIV
[SCSI] stex: use resid for xfer len information
[SCSI] Add Brownie 1200U3P to blacklist
[SCSI] scsi.c: convert to use the data buffer accessors
...
This patch tidies up scsi_add_lun a bit. I rewrote the kerneldoc to match
the actual parameters, moved the check for RBC and MMC REPORT_LUN devices
away from the switch(), changed the setup of sdev->type to account for
BLIST_ISROM, moved the check for BLIST_NO_ULD_ATTACH further down in
the function, removed a bogus comment and fixed some whitespace issues.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
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>
CONFIG_MODULES=y
CONFIG_SCSI=y
CONFIG_SCSI_SCAN_ASYNC=y
CONFIG_SCSI_WAIT_SCAN=m
2.6.21-rc5-mm2 VFS panics unable to find my root on /dev/sda2, but boots
okay if I change drivers/scsi/Kconfig to "default y" instead of "default m"
for SCSI_WAIT_SCAN.
Make sure there's a late_initcall to scsi_complete_async_scans when it's
built in, so a monolithic SCSI_SCAN_ASYNC kernel can rely on the scans
being completed before trying to mount root, even if they're slow.
[akpm@linux-foundation.org: build fixes]
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The #ifdef MODULE around the export of scsi_complete_async_scans()
which is the API the scsi_wait_scan module uses is incorrect and
causes the symbol to be undefined in certain circumstances leading to
a build failure. Remove the defines.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6:
[SCSI] SCSI core: better initialization for sdev->scsi_level
[SCSI] scsi_proc.c: display sdev->scsi_level correctly
[SCSI] megaraid_sas: update version and author info
[SCSI] megaraid_sas: return sync cache call with success
[SCSI] megaraid_sas: replace pci_alloc_consitent with dma_alloc_coherent in ioctl path
[SCSI] megaraid_sas: add bios_param in scsi_host_template
[SCSI] megaraid_sas: do not process cmds if hw_crit_error is set
[SCSI] scsi_transport.h should include scsi_device.h
[SCSI] aic79xx: remove extra newline from info message
[SCSI] scsi_scan.c: handle bad inquiry responses
[SCSI] aic94xx: tie driver to the major number of the sequencer firmware
[SCSI] lpfc: add PCI error recovery support
[SCSI] megaraid: pci_module_init to pci_register_driver
[SCSI] tgt: fix the user/kernel ring buffer interface
[SCSI] sgiwd93: interfacing to wd33c93
[SCSI] wd33c93: Fast SCSI with WD33C93B
This patch will affect the CDB in INQUIRY commands sent to LUNs above 0
when LUN-0 reports a scsi_level of 0; the LUN bits will no longer be set
in the second byte of the CDB. This is as it should be. Nevertheless,
it's possible that some wacky device might be adversely affected. I doubt
anyone will complain...
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
A particular USB device has been reporting short inquiry lengths. The
SCSI code cannot operate properly unless we get an inquiry length of
36 or above (because of the way we parse vendor and product), so
assume at least 36 bytes are valid even if the device reports fewer.
This is wrong, but it's no worse than what we're doing now (using the
garbage beyond the last reported valid byte).
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
I had thought that all drivers which didn't call scsi_scan_host()
called scsi_scan_target(). Some, such as sbp2, mptsas and libata-scsi,
call scsi_add_device() or __scsi_add_device(). We just need to wait
for the currently executing async scans to complete first. This is the
same code that's in scsi_scan_target(), except that we have to return
an error instead of void when we're declining to scan at all.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Hi,
Minor typo ...
In my first iteration of patches (that got merged), the
BLIST_ATTACH_PQ3 actually had the value 0x800000, but that
got changed later to avoid conflicts. This piece must have
been overlooked.
You could obviously do something like %x and then add the
bitflags, but that looks overkill for something that does
not tend to change.
Please merge.
(Patch applied against latest 2.6.20rc version that I tested.)
From: Kurt Garloff <kurt@garloff.de>
Subject: [SCSI SCAN] Fix logging message for PQ3 devices
The blacklist flags BLIST_ATTACH_PQ3 has value 0x1000000,
not 0x800000.
Signed-off-by: Kurt Garloff <garloff@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Apparently no ATAPI CD/DVD actually supports REPORT LUNS (in spite of
claiming scsi-3 compliance, where it's mandatory) and worse, some
crash or flake out on being sent the command. This may actually be
due to a conflict between SPC and MMC with MMC not listing REPORT LUNS
as mandatory. The same standards conflict exists for RBC as well.
Fix all of this by reversing the blacklists for CDROM and RBC devices
(i.e. now they have to have the BLIST_REPORTLUNS2 flag set even if the
inquiry data returns scsi-3 compliance).
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
If either scsi_complete_async_scans() is called a second time
before the first call has finished, or a host scan is started while
scsi_complete_async_scans() is still sleeping, it would fail to wake up
the other task, which would sleep forever.
I've changed the kernel-doc to make it clear that
scsi_complete_async_scans() only guarantees that scans which started
before it was called are guaranteed to have finished when it returns.
I considered making it wait until all scans are completed, but it can't
guarantee that no more scans will start before it returns anyway, and it
runs the risk of confusing other callers of scsi_complete_async_scans()
for hosts actually scanning.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Conflicts:
drivers/pcmcia/ds.c
Fix up merge failures with Linus's head and fix new compile failures.
Signed-Off-By: David Howells <dhowells@redhat.com>
If a driver can find its own targets, it can now fill in scan_finished and
(optionally) scan_start in the scsi_host_template. Then, when it calls
scsi_scan_host(), it will be called back (from a thread if asynchronous
discovery is enabled), first to start the scan, and then at intervals to
check if the scan is completed.
Also make scsi_prep_async_scan and scsi_finish_async_scan static.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Drivers that called scsi_scan_target() instead of scsi_scan_host() were
still adding devices; this needs to be under the control of userspace,
not the driver.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Without this patch, the user has to add a kernel command line parameter
to get asynchronous SCSI scanning. Now they can select the default at
compile time and still override it at boot time if they need to.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Pass the work_struct pointer to the work function rather than context data.
The work function can use container_of() to work out the data.
For the cases where the container of the work_struct may go away the moment the
pending bit is cleared, it is made possible to defer the release of the
structure by deferring the clearing of the pending bit.
To make this work, an extra flag is introduced into the management side of the
work_struct. This governs auto-release of the structure upon execution.
Ordinarily, the work queue executor would release the work_struct for further
scheduling or deallocation by clearing the pending bit prior to jumping to the
work function. This means that, unless the driver makes some guarantee itself
that the work_struct won't go away, the work function may not access anything
else in the work_struct or its container lest they be deallocated.. This is a
problem if the auxiliary data is taken away (as done by the last patch).
However, if the pending bit is *not* cleared before jumping to the work
function, then the work function *may* access the work_struct and its container
with no problems. But then the work function must itself release the
work_struct by calling work_release().
In most cases, automatic release is fine, so this is the default. Special
initiators exist for the non-auto-release case (ending in _NAR).
Signed-Off-By: David Howells <dhowells@redhat.com>
This patch (as810c) copies a minimum of 36 bytes of INQUIRY data, even if
the device claims that not all of them are valid. Often badly behaved
devices put plausible data in the Vendor, Product, and Revision strings but
set the Additional Length byte to a small value. Using potentially valid
data is certainly better than allocating a short buffer and then reading
beyond the end of it, which is what we do now.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Since it often takes around 20-30 seconds to scan a scsi bus, it's
highly advantageous to do this in parallel with other things. The bulk
of this patch is ensuring that devices don't change numbering, and that
all devices are discovered prior to trying to start init. For those
who build SCSI as modules, there's a new scsi_wait_scan module that will
ensure all bus scans are finished.
This patch only handles drivers which call scsi_scan_host. Fibre Channel,
SAS, SATA, USB and Firewire all need additional work.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Spotted by: Dan Aloni <da-xx@monatomic.org>
The problem is there's inconsistent locking semantic usage of
scsi_alloc_target(). Two callers assume the target comes back with
reference unincremented and the third assumes its incremented. Fix by
always making the reference incremented on return. Also fix path in
target alloc that could consistently increment the parent lock.
Finally document scsi_alloc_target() so its callers know what the
expectations are.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Sanitize the Vendor, Product, and Revision strings contained in an
INQUIRY result by setting all non-graphic or non-ASCII characters to ' '.
Since the standard disallows such characters, this will affect
only non-compliant devices.
To help maintain backward compatibility, NUL characters are treated
specially. They are taken as string terminators; they and all the
following characters are set to ' '. If some valid characters get
erased as a result... well, we weren't seeing them before so we haven't
lost anything.
The primary purpose of this change is to allow blacklist entries to
match devices with illegal Vendor or Product strings.
In addition, the patch updates a couple of function prototypes, giving
inq_result its correct type (unsigned char *).
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Some targets may return slight variations of PQ and PDT to indicate
no LUN mapped. USB UFI setting PDT=0x1f but having reserved bits for
PQ is one example, and NetApp targets returning PQ=1 and PDT=0x1f is
another. Both instances seem like reasonable responses according to
SPC-3 and UFI specs.
The current scsi_probe_and_add_lun() code adds a scsi_device
for targets that return PQ=1 and PDT=0x1f. This causes LUNs of type
"UNKNOWN" to show up in /proc/scsi/scsi when no LUNs are mapped.
In addition, subsequent rescans fail to recognize LUNs that may be
added on the target, unless preceded by a write to the delete attribute
of the "UNKNOWN" LUN.
This patch addresses this problem by skipping over the scsi_add_lun()
when PQ=1,PDT=0x1f is encountered, and just returns
SCSI_SCAN_TARGET_PRESENT.
Signed-off-by: Dave Wysochanski <davidw@netapp.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
A recent drivers base commit:
3e95637a48
Caused the bus to be added to dev_printk, so now our SCSI inquiry short
messages print like this:
scsiscsi 2:0:0:0: Direct access IBM-ESXS ST973401SS B519 PQ: 0 ANSI: 5
Just remove the "scsi" from the sdev_printk to compensate.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- Replace scsi_device_types array API with scsi_device_type function API.
Gets rid of a lot of common code, as well as being easier to use.
- Add the new device types in SPC4 r05a, and rename some of the older ones.
- Reformat the printing of inquiry data; now fits on one line and
includes PQ.
I think I've addressed all the feedback from the previous versions. My
current test box prints:
scsi 2:0:1:0: Direct access HP 18.2G ATLAS10K3_18_SCA HP05 PQ: 0 ANSI: 2
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Conflicts:
drivers/scsi/nsp32.c
drivers/scsi/pcmcia/nsp_cs.c
Removal of randomness flag conflicts with SA_ -> IRQF_ global
replacement.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
If a device gets offlined as a result of the Inquiry sent
during scanning, the following oops can occur. After the
disk gets put into the SDEV_OFFLINE state, the error handler
sends back the failed inquiry, which wakes the thread doing
the scan. This starts a race between the scanning thread
freeing the scsi device and the error handler calling
scsi_run_host_queues to restart the host. Since the disk
is in the SDEV_OFFLINE state, scsi_device_get will still
work, which results in __scsi_iterate_devices getting
a reference to the scsi disk when it shouldn't.
The following execution thread causes the oops:
CPU 0 (scan) CPU 1 (eh)
---------------------------------------------------------
scsi_probe_and_add_lun
....
scsi_eh_offline_sdevs
scsi_eh_flush_done_q
scsi_destroy_sdev
scsi_device_dev_release
scsi_restart_operations
scsi_run_host_queues
__scsi_iterate_devices
get_device
scsi_device_dev_release_usercontext
scsi_run_queue
<---OOPS--->
The patch fixes this by changing the state of the sdev to SDEV_DEL
before doing the final put_device, which should prevent the race
from occurring.
Original oops follows:
Badness in kref_get at lib/kref.c:32
Call Trace:
[C00000002F4476D0] [C00000000000EE20] .show_stack+0x68/0x1b0 (unreliable)
[C00000002F447770] [C00000000037515C] .program_check_exception+0x1cc/0x5a8
[C00000002F447840] [C00000000000446C] program_check_common+0xec/0x100
Exception: 700 at .kref_get+0x10/0x28
LR = .kobject_get+0x20/0x3c
[C00000002F447B30] [C00000002F447BC0] 0xc00000002f447bc0 (unreliable)
[C00000002F447BB0] [C000000000254BDC] .get_device+0x20/0x3c
[C00000002F447C30] [D000000000063188] .scsi_device_get+0x34/0xdc [scsi_mod]
[C00000002F447CC0] [D0000000000633EC] .__scsi_iterate_devices+0x50/0xbc [scsi_mod]
[C00000002F447D60] [D00000000006A910] .scsi_run_host_queues+0x34/0x5c [scsi_mod]
[C00000002F447DF0] [D000000000069054] .scsi_error_handler+0xdb4/0xe44 [scsi_mod]
[C00000002F447EE0] [C00000000007B4E0] .kthread+0x128/0x178
[C00000002F447F90] [C000000000025E84] .kernel_thread+0x4c/0x68
Unable to handle kernel paging request for <7>PCI: Enabling device: (0002:41:01.1), cmd 143
data at address 0x000001b8
Faulting instruction address: 0xd0000000000698e4
sym1: <1010-66> rev 0x1 at pci 0002:41:01.1 irq 216
sym1: No NVRAM, ID 7, Fast-80, LVD, parity checking
sym1: SCSI BUS has been reset.
scsi2 : sym-2.2.2
cpu 0x0: Vector: 300 (Data Access) at [c00000002f447a30]
pc: d0000000000698e4: .scsi_run_queue+0x2c/0x218 [scsi_mod]
lr: d00000000006a904: .scsi_run_host_queues+0x28/0x5c [scsi_mod]
sp: c00000002f447cb0
msr: 9000000000009032
dar: 1b8
dsisr: 40000000
current = 0xc0000000045fecd0
paca = 0xc00000000048ee80
pid = 1123, comm = scsi_eh_1
enter ? for help
[c00000002f447d60] d00000000006a904 .scsi_run_host_queues+0x28/0x5c [scsi_mod]
[c00000002f447df0] d000000000069054 .scsi_error_handler+0xdb4/0xe44 [scsi_mod]
[c00000002f447ee0] c00000000007b4e0 .kthread+0x128/0x178
[c00000002f447f90] c000000000025e84 .kernel_thread+0x4c/0x68
Signed-off-by: Brian King <brking@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
With Achim patch the last user (gdth) is switched away from scsi_request
so we an kill it now. Also disables some code in i2o_scsi that was
broken since the sg driver stopped using scsi_requests.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Mike Christie