linux-toradex.git/drivers/scsi/scsi.c, branch v4.13-rc5

scsi: default to scsi-mq

2017-06-28T01:31:12+00:00

Remove the SCSI_MQ_DEFAULT config option and default to the blk-mq I/O
path now that we had plenty of testing, and have I/O schedulers for
blk-mq.  The module option to disable the blk-mq path is kept around for
now.

Signed-off-by: Christoph Hellwig 
Signed-off-by: Martin K. Petersen

scsi: Only add commands to the device command list if required by the LLD

2017-06-13T00:55:59+00:00

Just like for the scsi-mq code path, in the single queue SCSI code path
only add commands to the per-device command list if required by the SCSI
LLD. This patch will make it easier to merge the single-queue and
multiqueue command initialization code.

Signed-off-by: Bart Van Assche 
Reviewed-by: Christoph Hellwig 
Reviewed-by: Hannes Reinecke 
Cc: Johannes Thumshirn 
Signed-off-by: Martin K. Petersen

scsi: Skip deleted devices in __scsi_device_lookup

2017-05-09T01:36:06+00:00

When a device is unplugged from a SCSI controller, if the scsi_device is
still in use by application layer, it won't get released until users
close it.

In this case, scsi_device_remove just set the scsi_device's state to be
SDEV_DEL. But if you plug the disk just before the old scsi_device is
released, then there will be two scsi_device structures in
scsi_host->__devices. When the next unplug event happens, some low-level
drivers will check whether the scsi_device has been added to host (for
example the MegaRAID SAS series controller) by calling
scsi_device_lookup(call __scsi_device_lookup) in function
megasas_aen_polling. __scsi_device_lookup will return the first
scsi_device. Because its state is SDEV_DEL, the scsi_device_lookup will
return NULL, making the low-level driver assume that the scsi_device has
been removed, and won't call scsi_device_remove which will lead to hot
swap failure.

Signed-off-by: Zhou Zhengping 
Tested-by: Zeng Rujia 
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=195607
Reviewed-by: Hannes Reinecke 
Signed-off-by: Martin K. Petersen

block: introduce blk_rq_is_passthrough

2017-01-31T21:00:34+00:00

This can be used to check for fs vs non-fs requests and basically
removes all knowledge of BLOCK_PC specific from the block layer,
as well as preparing for removing the cmd_type field in struct request.

Signed-off-by: Christoph Hellwig 
Signed-off-by: Jens Axboe

scsi: allocate scsi_cmnd structures as part of struct request

2017-01-27T22:08:35+00:00

Rely on the new block layer functionality to allocate additional driver
specific data behind struct request instead of implementing it in SCSI
itѕelf.

Signed-off-by: Christoph Hellwig 
Acked-by: Martin K. Petersen 
Reviewed-by: Hannes Reinecke 
Signed-off-by: Jens Axboe

scsi: remove scsi_cmd_dma_pool

2017-01-27T22:08:35+00:00

There is no need for GFP_DMA allocations of the scsi_cmnd structures
themselves, all that might be DMAed to or from is the actual payload,
or the sense buffers.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Hannes Reinecke 
Acked-by: Martin K. Petersen 
Signed-off-by: Jens Axboe

scsi: respect unchecked_isa_dma for blk-mq

2017-01-27T22:08:35+00:00

Currently blk-mq always allocates the sense buffer using normal GFP_KERNEL
allocation.  Refactor the cmd pool code to split the cmd and sense allocation
and share the code to allocate the sense buffers as well as the sense buffer
slab caches between the legacy and blk-mq path.

Note that this switches to lazy allocation of the sense slab caches - the
slab caches (not the actual allocations) won't be destroy until the scsi
module is unloaded instead of keeping track of hosts using them.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Hannes Reinecke 
Acked-by: Martin K. Petersen 
Signed-off-by: Jens Axboe

scsi: remove gfp_flags member in scsi_host_cmd_pool

2017-01-27T22:08:35+00:00

When using the slab allocator we already decide at cache creation time if
an allocation comes from a GFP_DMA pool using the SLAB_CACHE_DMA flag,
and there is no point passing the kmalloc-family only GFP_DMA flag to
kmem_cache_alloc.  Drop all the infrastructure for doing so.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Johannes Thumshirn 
Acked-by: Martin K. Petersen 
Signed-off-by: Jens Axboe

block: add code to track actual device queue depth

2016-11-05T23:09:53+00:00

For blk-mq, ->nr_requests does track queue depth, at least at init
time. But for the older queue paths, it's simply a soft setting.
On top of that, it's generally larger than the hardware setting
on purpose, to allow backup of requests for merging.

Fill a hole in struct request with a 'queue_depth' member, that
drivers can call to more closely inform the block layer of the
real queue depth.

Signed-off-by: Jens Axboe 
Reviewed-by: Jan Kara

scsi: Avoid that toggling use_blk_mq triggers a memory leak

2016-09-27T00:58:42+00:00

This patch avoids that the following memory leak is triggered if
use_blk_mq is disabled after a SCSI host has been allocated by the
ib_srp driver and before the same SCSI host is freed:

unreferenced object 0xffff8803a168c568 (size 256):
  backtrace:
    [] kmemleak_alloc+0x45/0xa0
    [] __kmalloc_node+0x1e4/0x400
    [] blk_mq_alloc_tag_set+0xb4/0x230
    [] scsi_mq_setup_tags+0xc7/0xd0
    [] scsi_add_host_with_dma+0x216/0x2d0
    [] srp_create_target+0xe55/0x13d0 [ib_srp]
    [] dev_attr_store+0x13/0x20
    [] sysfs_kf_write+0x40/0x50
    [] kernfs_fop_write+0x137/0x1c0
    [] __vfs_write+0x23/0x140
    [] vfs_write+0xb0/0x190
    [] SyS_write+0x44/0xa0
    [] entry_SYSCALL_64_fastpath+0x18/0xa8

Fixes: 9aa9cc4221f5 ("scsi: remove the disable_blk_mq host flag")
Signed-off-by: Bart Van Assche 
Cc: Christoph Hellwig 
Cc: Martin K. Petersen 
Cc: 
Reviewed-by: Christoph Hellwig 
Signed-off-by: Martin K. Petersen