Skip to contentSkip to conversation

Block driver changes for 3.20

GIT PULL Jens Axboe 4 participants 9 messages Feb 12, 2015 – Feb 13, 2015 libata

lore
Share & tools
lore

Jens Axboe

author
#0 lore raw

Hi Linus,

This is the pull request for block drivers for 3.20. This branch sits on top of (an earlier) for-3.20/core. It contains:

- The 4k/partition fixes for brd from Boaz/Matthew.

- A few xen front/back block fixes from David Vrabel and Roger Pau Monne.

- Floppy changes from Takashi, cleaning the device file creation.

- Switching libata to use the new blk-mq tagging policy, removing code (and a suboptimal implementation) from libata. This will throw you a merge conflict, since a bug in the original libata tagging code was fixed since this code was branched. Trivial. From Shaohua.

- Conversion of loop to blk-mq, from Ming Lei.

- Cleanup of the io_schedule() handling in bsg from Peter Zijlstra. He claims it improves on unreadable code, which will cost him a beer.

- Maintainer update or NDB, now handled by Markus Pargmann.

- NVMe:
- Optimization from me that avoids a kmalloc/kfree per IO for smaller (<= 8KB) IO. This cuts about 1% of high IOPS CPU overhead.
- Removal of (now) dead RCU code, a relic from before NVMe was converted to blk-mq.

Please pull!

git://git.kernel.dk/linux-block.git for-3.20/drivers

----------------------------------------------------------------
Boaz Harrosh (2):
brd: Fix all partitions BUGs
brd: Request from fdisk 4k alignment

David Vrabel (1):
xen-blkback: default to X86_32 ABI on x86

Jens Axboe (5):
loop: add blk-mq.h include
Merge branch 'for-3.20/core' into for-3.20/drivers
null_blk: suppress invalid partition info Merge branch 'for-3.20/core' into for-3.20/drivers
NVMe: avoid kmalloc/kfree for smaller IO

Markus Pargmann (1):
MAINTAINERS: Update NBD maintainer

Matthew Wilcox (1):
axonram: Fix bug in direct_access

Ming Lei (4):
block: loop: improve performance via blk-mq
block: loop: say goodby to bio
block: loop: introduce lo_discard() and lo_req_flush()
block: loop: don't handle REQ_FUA explicitly

Peter Zijlstra (1):
block: Simplify bsg complete all

Roger Pau Monne (2):
xen-blkback,xen-blkfront: add myself as maintainer
xen-blkfront: fix accounting of reqs when migrating

Shaohua Li (3):
libata: use blk taging
libata: move sas ata tag allocation to libata-scsi.c
libata: make sata_sil24 use fifo tag allocator

Takashi Iwai (1):
floppy: Avoid manual call of device_create_file()

kaoudis (1):
NVMe: within nvme_free_queues(), delete RCU sychro/deferred free

 MAINTAINERS                        |   4 +-
 arch/powerpc/sysdev/axonram.c      |   2 +-
 block/bsg.c                        |  72 +++----
 drivers/ata/libata-core.c          |  67 ++----
 drivers/ata/libata-scsi.c          |  30 ++-
 drivers/ata/libata.h               |   4 +-
 drivers/ata/sata_sil24.c           |   1 +
 drivers/block/brd.c                | 109 +++++-----
 drivers/block/floppy.c             |  17 +-
 drivers/block/loop.c               | 416 ++++++++++++++++++-------------------
 drivers/block/loop.h               |  18 +-
 drivers/block/null_blk.c           |   2 +-
 drivers/block/nvme-core.c          | 128 ++++++++----
 drivers/block/xen-blkback/common.h |   9 +
 drivers/block/xen-blkback/xenbus.c |   4 +-
 drivers/block/xen-blkfront.c       |   4 +-
 include/linux/libata.h             |   5 +-
 include/linux/nvme.h               |   3 +-
 include/linux/wait.h               |  15 ++
 19 files changed, 473 insertions(+), 437 deletions(-)
Signature · 1 line
Jens Axboe
#1 lore raw
Quoted 5 lines · Jens Axboe

- Switching libata to use the new blk-mq tagging policy, removing code (and a suboptimal implementation) from libata. This will throw you a merge conflict, since a bug in the original libata tagging code was fixed since this code was branched. Trivial. From Shaohua.

Somebody should still check my resolution, since the code had been moved to another file. Tejun / Dan, it's your commit 72dd299d5039 ("libata: allow sata_sil24 to opt-out of tag ordered submission") that added a ATA_FLAG_LOWTAG, which conflicts with commit 98bd4be1ba95 ("libata: move sas ata tag allocation to libata-scsi.c") that moved the code.

I verified that it all looks sane, and still compiles, but somebody should verify that the tag allocation changes still *work*, particularly for that sata_sil24 case.

(I haven't pushed out quite yet, I'm waiting for the rest to compile cleanly too, and then I'll do a local compile and boot to see that it all works for me, but I'm writing this as a heads-up)

Linus

#8 lore raw
In reply to Jens Axboe

Hi Linus,

Quoted 3 lines · On Thu, 12 Feb 2015 13:57:52 -0700 Jens Axboe <[email protected]> wrote:

Shaohua Li (3):
libata: move sas ata tag allocation to libata-scsi.c

So, I assume you figured out the attached semantic conflict ...

Signature · 2 lines
Cheers,
Stephen Rothwell                    [email protected]

Hi Jens,

Today's linux-next merge of the block tree got a conflict in drivers/ata/libata-core.c between commit 72dd299d5039 ("libata: allow sata_sil24 to opt-out of tag ordered submission") from Linus' tree and commit 98bd4be1ba95 ("libata: move sas ata tag allocation to libata-scsi.c") from the block tree.

I fixed it up (I used the version from the block tree and then added the following patch) and can carry the fix as necessary (no action is required).

From: Stephen Rothwell <[email protected]>
Date: Tue, 27 Jan 2015 14:59:51 +1100
Subject: [PATCH] libata: fix for move of sas tag allocation code

---
 drivers/ata/libata-scsi.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/ata/libata-scsi.c +4 -1
index fd9be1756f0d..055e75d2176e 100644
@@ -4240,7 +4240,10 @@ int ata_sas_allocate_tag(struct ata_port *ap)
 	unsigned int i, tag;
 
 	for (i = 0, tag = ap->sas_last_tag + 1; i < max_queue; i++, tag++) {
-		tag = tag < max_queue ? tag : 0;
+		if (ap->flags & ATA_FLAG_LOWTAG)
+			tag = i;
+		else
+			tag = tag < max_queue ? tag : 0;
 
 		/* the last tag is reserved for internal command. */
 		if (tag == ATA_TAG_INTERNAL)
Signature · 5 lines
2.1.4

-- 
Cheers,
Stephen Rothwell                    [email protected]
author
#2 lore raw
In reply to Linus Torvalds
Quoted 14 lines · Linus Torvalds
Quoted 5 lines · Jens Axboe

- Switching libata to use the new blk-mq tagging policy, removing code (and a suboptimal implementation) from libata. This will throw you a merge conflict, since a bug in the original libata tagging code was fixed since this code was branched. Trivial. From Shaohua.

Somebody should still check my resolution, since the code had been moved to another file. Tejun / Dan, it's your commit 72dd299d5039 ("libata: allow sata_sil24 to opt-out of tag ordered submission") that added a ATA_FLAG_LOWTAG, which conflicts with commit 98bd4be1ba95 ("libata: move sas ata tag allocation to libata-scsi.c") that moved the code.

Quoted 7 lines

I verified that it all looks sane, and still compiles, but somebody should verify that the tag allocation changes still *work*, particularly for that sata_sil24 case.

(I haven't pushed out quite yet, I'm waiting for the rest to compile cleanly too, and then I'll do a local compile and boot to see that it all works for me, but I'm writing this as a heads-up)

When the code was moved, it also morphed into the sas specific tag allocation. And the LOWTAG part was for sil24, which doesn't call that code anymore. So either resolution is fine:

1) Drop the LOWTAG part of ata_sas_allocate_tag(), which is the easy resolve. Or,
2) Add the two LOWTAG lines in ata_sas_allocate_tag().

So don't worry too much about it, if it compiles, it's good...

Signature · 1 line
Jens Axboe
#3 lore raw
Quoted 4 lines · Jens Axboe

When the code was moved, it also morphed into the sas specific tag allocation. And the LOWTAG part was for sil24, which doesn't call that code anymore.

Hmm. So where does the sil24 get its tag allocation from then? Because then *that* code presumably needs the LOWTAG hackery..

Linus

author Acked-by Tejun Heo Acked-by Dan Williams
#4 lore raw
Quoted 8 lines · Linus Torvalds
Quoted 4 lines · Jens Axboe

When the code was moved, it also morphed into the sas specific tag allocation. And the LOWTAG part was for sil24, which doesn't call that code anymore.

Hmm. So where does the sil24 get its tag allocation from then? Because then *that* code presumably needs the LOWTAG hackery..

commit 9269e23496ddd83b974536019cf420520df0ee7f

Author: Shaohua Li <[email protected]>
Date:   Fri Jan 23 20:17:59 2015 -0800

libata: make sata_sil24 use fifo tag allocator

libata starts using block tag now, we can use BLK_TAG_ALLOC_FIFO to solve the sata_sil24 tag bug.

https://bugzilla.kernel.org/show_bug.cgi?id=87101

Signature · 1 line
Jens Axboe
#5 lore raw
Quoted 9 lines · Jens Axboe

commit 9269e23496ddd83b974536019cf420520df0ee7f

Author: Shaohua Li <[email protected]>
Date:   Fri Jan 23 20:17:59 2015 -0800

libata: make sata_sil24 use fifo tag allocator

libata starts using block tag now, we can use BLK_TAG_ALLOC_FIFO to solve the sata_sil24 tag bug.

Ahh, ok. So we should probably just remove the now unused/useless ATA_FLAG_LOWTAG entirely.

I'll leave my resolution alone, and leave that for Tejun if he cares.

Thanks,

Linus

author
#6 lore raw
Quoted 6 lines · Linus Torvalds
Quoted 5 lines · Jens Axboe

commit 9269e23496ddd83b974536019cf420520df0ee7f

Author: Shaohua Li <[email protected]>
Date:   Fri Jan 23 20:17:59 2015 -0800
Quoted 7 lines
Quoted 4 lines

libata: make sata_sil24 use fifo tag allocator

libata starts using block tag now, we can use BLK_TAG_ALLOC_FIFO to solve the sata_sil24 tag bug.

Ahh, ok. So we should probably just remove the now unused/useless ATA_FLAG_LOWTAG entirely.

Yeah, we can kill that after the fact. I probably should have pulled in that commit from Dan and fixed this up... But in any case, it'll be fine as is right now.

Signature · 1 line
Jens Axboe
#7 lore raw

Hello,

Quoted 3 lines · Jens Axboe

Yeah, we can kill that after the fact. I probably should have pulled in that commit from Dan and fixed this up... But in any case, it'll be fine as is right now.

Yeap, the current code should be fine. I'll remove the LOWTAG flag from libata side in the coming devel window.

Thanks.

Signature · 1 line
tejun