linux.git/drivers/block, branch v2.6.31

aoe: allocate unused request_queue for sysfs

2009-09-09T12:10:18+00:00

Andy Whitcroft reported an oops in aoe triggered by use of an
incorrectly initialised request_queue object:

  [ 2645.959090] kobject '' (ffff880059ca22c0): tried to add
		an uninitialized object, something is seriously wrong.
  [ 2645.959104] Pid: 6, comm: events/0 Not tainted 2.6.31-5-generic #24-Ubuntu
  [ 2645.959107] Call Trace:
  [ 2645.959139] [] kobject_add+0x5f/0x70
  [ 2645.959151] [] blk_register_queue+0x8b/0xf0
  [ 2645.959155] [] add_disk+0x8f/0x160
  [ 2645.959161] [] aoeblk_gdalloc+0x164/0x1c0 [aoe]

The request queue of an aoe device is not used but can be allocated in
code that does not sleep.

Bruno bisected this regression down to

  cd43e26f071524647e660706b784ebcbefbd2e44

  block: Expose stacked device queues in sysfs

"This seems to generate /sys/block/$device/queue and its contents for
 everyone who is using queues, not just for those queues that have a
 non-NULL queue->request_fn."

Addresses http://bugs.launchpad.net/bugs/410198
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=13942

Note that embedding a queue inside another object has always been
an illegal construct, since the queues are reference counted and
must persist until the last reference is dropped. So aoe was
always buggy in this respect (Jens).

Signed-off-by: Ed Cashin 
Cc: Andy Whitcroft 
Cc: "Rafael J. Wysocki" 
Cc: Bruno Premont 
Cc: Martin K. Petersen 
Cc: Andrew Morton 
Signed-off-by: Jens Axboe

mg_disk: Add missing ready status check on mg_write()

2009-07-28T06:57:33+00:00

When last sector is written, ready bit of status register should be
checked.

Signed-off-by: unsik Kim 
Acked-by: Bartlomiej Zolnierkiewicz 
Signed-off-by: Jens Axboe

mg_disk: fix issue with data integrity on error in mg_write()

2009-07-28T06:56:34+00:00

We cannot acknowledge the sector write before checking its status
(which is done on the next loop iteration) and we also need to do
the final status register check after writing the last sector.

Fix mg_write() to match mg_write_intr() in this regard.

While at it:
- add mg_read_one() and mg_write_one() helpers
- always use MG_SECTOR_SIZE and remove MG_STORAGE_BUFFER_SIZE

[bart: thanks to Tejun for porting the patch over recent block changes]

Cc: unsik Kim 
Cc: Tejun Heo 
Signed-off-by: Bartlomiej Zolnierkiewicz 

===================================================================
Signed-off-by: Jens Axboe

mg_disk: fix reading invalid status when use polling driver

2009-07-28T06:52:07+00:00

When using polling driver, little delay is required to access
status register. Without this, host might read invalid status.

Signed-off-by: unsik Kim 
Signed-off-by: Jens Axboe

mg_disk: remove prohibited sleep operation

2009-07-28T06:52:06+00:00

mflash's polling driver operate in standard request_fn_proc's context,
sleep in this isn't permitted.

Signed-off-by: unsik Kim 
Signed-off-by: Jens Axboe

Merge branch 'tj-block-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/misc

2009-07-22T17:06:33+00:00

* 'tj-block-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/misc:
  virtio_blk: mark virtio_blk with __refdata to kill spurious section mismatch
  block: sysfs fix mismatched queue_var_{store,show} in 64bit kernel
  ataflop: adjust NULL test
  block: fix failfast merge testing in elv_rq_merge_ok()
  z2ram: Small cleanup for z2ram.c

virtio_blk: mark virtio_blk with __refdata to kill spurious section mismatch

2009-07-19T01:46:48+00:00

The variable virtio_blk references the function virtblk_probe() (which
is in .devinit section) and also references the function
virtblk_remove() ( which is in .devexit section). So, virtio_blk
simultaneously refers .devinit and .devexit section. To avoid this
messup, we mark virtio_blk as __refdata.

We were warned by the following warning:

  LD      drivers/block/built-in.o
  WARNING: drivers/block/built-in.o(.data+0xc8dc): Section mismatch in
  reference from the variable virtio_blk to the function
  .devinit.text:virtblk_probe()
  The variable virtio_blk references
  the function __devinit virtblk_probe()
  If the reference is valid then annotate the
  variable with __init* or __refdata (see linux/init.h) or name the variable:
  *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console,

  WARNING: drivers/block/built-in.o(.data+0xc8e0): Section mismatch in
  reference from the variable virtio_blk to the function
  .devexit.text:virtblk_remove()
  The variable virtio_blk references
  the function __devexit virtblk_remove()
  If the reference is valid then annotate the
  variable with __exit* (see linux/init.h) or name the variable:
  *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console,

Signed-off-by: Rakib Mullick 
Signed-off-by: Tejun Heo

virtio_blk: ioctl return value fix

2009-07-17T12:17:46+00:00

Block driver ioctl methods must return ENOTTY and not -ENOIOCTLCMD if
they expect the block layer to handle generic ioctls.

This triggered a BLKROSET failure in xfsqa #200.

Signed-off-by: Christoph Hellwig 
Signed-off-by: Rusty Russell

virtio_blk: don't bounce highmem requests

2009-07-17T12:17:46+00:00

By default a block driver bounces highmem requests, but virtio-blk is
perfectly fine with any request that fit into it's 64 bit addressing scheme,
mapped in the kernel virtual space or not.

Besides improving performance on highmem systems this also makes the
reproducible oops in __bounce_end_io go away (but hiding the real cause).

Signed-off-by: Christoph Hellwig 
Signed-off-by: Rusty Russell

ataflop: adjust NULL test

2009-07-17T06:29:58+00:00

dtp is derefenced on the lines above the test !dtp, and so it cannot be
NULL at this point.

A simplified version of the semantic match that finds this problem is as
follows: (http://www.emn.fr/x-info/coccinelle/)

// 
@r@
expression x,E,E1;
identifier f,l;
position p1,p2;
@@

*x@p1->f = E1;
... when != x = E
    when != goto l;
(
*x@p2 == NULL
|
*x@p2 != NULL
)
// 

Signed-off-by: Julia Lawall 
Signed-off-by: Tejun Heo