- 17 Feb, 2015 1 commit
-
-
Tony Battersby authored
When using the write()/read() interface for submitting commands, the SCSI generic driver does not call blk_put_request() on a completed SCSI command until userspace calls read() to get the command completion. Since scsi-mq uses a fixed number of preallocated requests, this makes it possible for userspace to exhaust the entire preallocated supply of requests. For places in the kernel that call blk_get_request() with GFP_KERNEL, this can cause the calling process to deadlock in a permanent unkillable I/O wait in blk_get_request() -> ... -> bt_get(). For places in the kernel that call blk_get_request() with GFP_ATOMIC, this can cause blk_get_request() always to return -EWOULDBLOCK. Note that these problems happen only if scsi-mq is enabled. Prevent the problems by calling blk_put_request() as soon as the SCSI command completes instead of waiting for userspace to call read(). Cc: <stable@vger.kernel.org> # 3.17+ Signed-off-by:
Tony Battersby <tonyb@cybernetics.com> Acked-by:
Douglas Gilbert <dgilbert@interlog.com> Tested-by:
James Bottomley <JBottomley@Parallels.com>
-
- 15 Feb, 2015 4 commits
-
-
Tony Battersby authored
Fix SCSI generic read() incorrectly returning success after detecting an error. Cc: <stable@vger.kernel.org> Signed-off-by:
-
Ondrej Zary authored
wd719x_template is missing the .module field, causing module refcount not to work, allowing to rmmod the driver while in use (mounted filesystem), causing an oops. Set .module to THIS_MODULE to fix the problem. Signed-off-by:
Ondrej Zary <linux@rainbow-software.org> Cc: <stable@vger.kernel.org> Signed-off-by:
-
Don Brace authored
Correct compiler warning introduced by hpsa-add-local-workqueue patch 6636e7f4 hpsa: Use local workqueues instead of system workqueues Suggested-by:
Kees Cook <keescook@chromium.org> Reviewed-by:
Scott Teel <scott.teel@pmcs.com> Reviewed-by:
Webb Scales <webbnh@hp.com> Signed-off-by:
Don Brace <don.brace@pmcs.com> Signed-off-by:
-
Minh Duc Tran authored
fixed invalid assignment of 64bit mask to host dma_boundary for scatter gather segment boundary limit. In reference to bug https://bugzilla.redhat.com/show_bug.cgi?id=1097141 Assert is seen with AMD cpu whenever calling pci_alloc_consistent. [ 29.406183] ------------[ cut here ]------------ [ 29.410505] kernel BUG at lib/iommu-helper.c:13! Signed-off-by:
Minh Tran <minh.tran@emulex.com> Fixes: 6733b39a Cc: <stable@vger.kernel.org> Signed-off-by:
-
- 06 Feb, 2015 1 commit
-
-
Robert Love authored
Acked-by:
Vasu Dev <vasu.dev@intel.com> Signed-off-by:
Robert Love <robert.w.love@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by:
-
- 05 Feb, 2015 1 commit
-
-
Hannes Reinecke authored
A shost_printk() statement was left over from debugging. It can safely be removed; the same information is displayed in the debugging message some lines further down. Cc: Ondrej Zary <linux@rainbow-software.org> Signed-off-by:
Hannes Reinecke <hare@suse.de> Signed-off-by:
Christoph Hellwig <hch@lst.de>
-
- 04 Feb, 2015 1 commit
-
-
Hannes Reinecke authored
dev_printk() is now a void function, so the related functions scmd_printk() and sdev_prefix_printk() should be made void, too. Reported-by:
Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by:
-
- 02 Feb, 2015 32 commits
-
-
Rasmus Villemoes authored
Using seq_putc to print a single character saves at least a strlen() call and a memory access, and may also give a small .text reduction. Signed-off-by:
Rasmus Villemoes <linux@rasmusvillemoes.dk> Reviewed-by:
Finn Thain <fthain@telegraphics.com.au> Signed-off-by:
-
Rasmus Villemoes authored
Consecutive seq_puts calls with literal strings may be replaced by a single call, saving a little .text. Signed-off-by:
-
Rasmus Villemoes authored
Using seq_printf to print a simple string is a lot more expensive than it needs to be, since seq_puts exists. Replace seq_printf with seq_puts when possible. Signed-off-by:
-
Rasmus Villemoes authored
Using seq_printf to print a simple string is a lot more expensive than it needs to be, since seq_puts exists. Replace seq_printf with seq_puts when possible. Signed-off-by:
-
Rasmus Villemoes authored
Using seq_printf to print a simple string is a lot more expensive than it needs to be, since seq_puts exists. Replace seq_printf with seq_puts when possible. Signed-off-by:
-
Rasmus Villemoes authored
The macro SPRINTF doesn't save a lot of typing or make the code more readable, and depending on a specific identifier (m) in the surrounding scope is generally frowned upon. Nuke it. Signed-off-by:
-
Bart Van Assche authored
The 'data_dir' variable is not used in sg_common_write(), hence remove this variable. Signed-off-by:
Bart Van Assche <bart.vanassche@sandisk.com> Acked-by:
-
Don Brace authored
Suggested-by:
Tomas Henzl <thenzl@redhat.com> Reviewed-by:
Kevin Barnett <Kevin.Barnett@pmcs.com> Signed-off-by:
-
Don Brace authored
Add in P840ar model name for gen9 Reviewed-by:
-
Don Brace authored
Add in gen9 controller model names Reviewed-by:
-
Robert Elliott authored
Detect failues when attempting to change controller to use simple or performant transport modes (mode change ack) rather than just proceeding ahead after timeouts. Return values are added to: hpsa_put_ctlr_into_performant_mode hpsa_wait_for_mode_change_ack and all their callers check/propagate the result. More consistency in printing errors and whether dev_err is used. Reviewed-by:
Robert Elliott <elliott@hp.com> Signed-off-by:
-
Robert Elliott authored
Shorten the wait for the CISS configuration table doorbell mode change acknowledgment from 300-600 s to 20 s, which is the value specified in the CISS specification that should be honored by all controllers. Wait using interruptible msleep() rather than uninterruptible usleep_range(), which triggers rt_sched timeout errors if the wait is long. Reviewed-by:
-
Webb Scales authored
Hoist the conditional out of do_not_scan_if_controller_locked_up() and place it in the caller (this improves the code structure, making it more consistent with other uses and enabling tail-call optimization); rename the function to hpsa_scan_complete(), and use it at the end of hpsa_scan_start() as well. Reviewed-by:
-
Webb Scales authored
Move the code which sets up the SG descriptor out of hpsa_scatter_gather() and into a subroutine where it can be reused (in the next patch). The Ext field is now assigned unconditionally: this makes the refactor much simpler, but more importantly it removes a conditional operation from inside the loop. The case for which the conditional formerly tested is now executed (unconditionally) after the loop is exited. Reviewed-by:
-
Stephen Cameron authored
Performance tweak, avoid unnecessary function calls. Reviewed-by:
-
Stephen Cameron authored
Printing the address of the command pointer is of little value, change to print the CDB. Reviewed-by:
-
Stephen Cameron authored
There's no reason for it to be a void *, it should be a struct scsi_cmnd * Reviewed-by:
-
Don Brace authored
Returning failed from the device reset handler will get the device kicked offline, which is fine if the controller is locked up anyhow. Cannot abort a command from a failed controller. Reviewed-by:
Justin Lindley <justin.lindley@pmcs.com> Signed-off-by:
-
Stephen Cameron authored
Command allocation is the thing that takes the longest in the main i/o path, so check for controller lockup immediately after this to prevent submitting commands to locked up controller as much as possible. Reviewed-by:
-
Stephen Cameron authored
In the code that translates logical drive LBAs to physical drive LBAs if we overflow the raid map disk data array we will get the wrong answers. We do not expect that to happen, but best to be on the safe side and guard against it anyway. Reviewed-by:
-
Stephen Cameron authored
Acking controller events on controllers that do not support it can cause such controllers to lock up. Reviewed-by:
Joe Handzik <joseph.t.handzik@hp.com> Signed-off-by:
-
Stephen Cameron authored
In set_encrypt_ioaccel2() and in hpsa_scsi_ioaccel_raid_map there were BUG_ONs that looked like this: BUG_ON(!(dev->offload_config && dev->offload_enabled)); But, In hpsa_ack_ctlr_events() we have this, /* Stop sending new RAID offload reqs via the IO accelerator */ scsi_block_requests(h->scsi_host); for (i = 0; i < h->ndevices; i++) h->dev[i]->offload_enabled = 0; hpsa_drain_accel_commands(h); So, we set offload_enabled = 0 for all drives, then do this drain_accel_commands, so that means accel commands could still be in flight, ie. perhaps having just been submitted into hpsa_scsi_ioaccel_raid_map concurrent with ->offload_enabled having just been set to zero. Reviewed-by:
-
Stephen Cameron authored
No need to check whether interrupt pending for MSI(X) and conversely, no need to check whether MSI(X) interrupts are being used when checking if interrupts are pending. Reviewed-by:
-
Don Brace authored
Reviewed-by:
-
Don Brace authored
Performance enhancement. Remove spin_locks from the driver. Reviewed-by:
-
Robert Elliott authored
Empirically, this improves performance slightly (~2% max IOPS) by allowing cmd_alloc to remember where it left off searching for free commands between calls instead of always starting its search at command 0. Reviewed-by:
-
Webb Scales authored
This means changing the allocator to reference count commands. The reference count is now the authoritative indicator of whether a command is allocated or not. The h->cmd_pool_bits bitmap is now only a heuristic hint to speed up the allocation process, it is no longer the authoritative record of allocated commands. Since we changed the command allocator to use reference counting as the authoritative indicator of whether a command is allocated, fail_all_outstanding_cmds needs to use the reference count not h->cmd_pool_bits for this purpose. Fix hpsa_drain_accel_commands to use the reference count as the authoritative indicator of whether a command is allocated instead of the h->cmd_pool_bits bitmap. Reviewed-by:
-
Don Brace authored
When using the ioaccel submission methods, requests destined for RAID volumes are sometimes diverted to physical devices. The OS has no or limited knowledge of these physical devices, so it is up to the driver to avoid pushing the device too hard. It is better to honor the physical device queue limit rather than making the device spew zillions of TASK SET FULL responses. This is so that hpsa based devices support /sys/block/sdNN/device/queue_type of simple, which lets the SCSI midlayer automatically adjust the queue_depth based on TASK SET FULL and GOOD status. Adjust the queue depth for a new device after it is created based on the maximum queue depths of the physical devices that constitute the device. This drops the maximum queue depth from .can_queue of 1024 to something like 174 for single-drive RAID-0, 348 for two-drive RAID-1, etc. It also adjusts for the ratio of data to parity drives. Reviewed-by:
-
Don Brace authored
Instead of kicking the commands all the way back to the mid layer, use a work queue. This enables having a mechanism for the driver to be able to resubmit the commands down the "normal" raid path without turning off the ioaccel feature entirely whenever an error is encountered on the ioaccel path, and prevent excessive rescanning of devices. Reviewed-by:
-
Stephen Cameron authored
Factor out the bottom part of the queuecommand function which is the part that builds commands for submitting down the "normal' RAID stack path of a Smart Array. Need to factor this out to improve how commands that were initially sent down one of the "ioaccellerated" paths but which have some sort of error condition are retried down the "normal" path. Reviewed-by:
-
Stephen Cameron authored
The original reasoning behind doing this was faulty. An error of some sort would be encountered, accelerated i/o would be disabled for that logical drive, the command would be kicked back out to the SCSI midlayer for a retry, and since i/o accelerator mode was disabled, it would get retried down the RAID path. However, something needs to turn ioaccellerator mode back on, and this rescan request was what did that. However, it was racy, and extremely bad for performance to rescan all devices, so, don't do that. Reviewed-by:
-
Don Brace authored
By not doing maintaining a list of queued commands, we can eliminate some spin locking in the main i/o path and gain significant improvement in IOPS. Remove the queuing code and the code that calls it; remove now-unused interrupt code; remove DIRECT_LOOKUP_BIT. Now that the passthru commands share the same command pool as the main i/o path, and the total size of the pool is less than or equal to the number of commands that will fit in the hardware fifo, there is no need to check to see if we are exceeding the hardware fifo's depth. Reviewed-by:
-
