Linux block layer
 help / color / mirror / Atom feed
From: Keith Busch <kbusch@meta.com>
To: <axboe@kernel.dk>, <linux-block@vger.kernel.org>,
	<martin.petersen@oracle.com>, <hch@lst.de>
Cc: <linux-nvme@lists.infradead.org>, Keith Busch <kbusch@kernel.org>
Subject: [PATCH] block: always allocate integrity buffer when required
Date: Thu, 8 May 2025 10:58:14 -0700	[thread overview]
Message-ID: <20250508175814.1176459-1-kbusch@meta.com> (raw)

From: Keith Busch <kbusch@kernel.org>

Many nvme metadata formats can not strip or generate the metadata on the
controller side. For these, a host provided integrity buffer is
mandatory, even if it isn't checked.

The block integrity read_verify and write_generate attributes prevent
allocating the metadata buffer, but we need it when the format requires
it, otherwise reads and writes will be rejected by the driver with IO
errors.

Add a new blk_integrity flag to indicate if the disk format requires an
integrity buffer. When this flag is set, provide an unchecked buffer if
the sysfs attributes disabled verify or generation. This fixes the
following nvme warning:

 ------------[ cut here ]------------
 WARNING: CPU: 1 PID: 371 at drivers/nvme/host/core.c:1036 nvme_setup_rw+0x122/0x210
 ...
 RIP: 0010:nvme_setup_rw+0x122/0x210
 ...
 Call Trace:
  <TASK>
  nvme_setup_cmd+0x1b4/0x280
  nvme_queue_rqs+0xc4/0x1f0 [nvme]
  blk_mq_dispatch_queue_requests+0x24a/0x430
  blk_mq_flush_plug_list+0x50/0x140
  __blk_flush_plug+0xc1/0x100
  __submit_bio+0x1c1/0x360
  ? submit_bio_noacct_nocheck+0x2d6/0x3c0
  submit_bio_noacct_nocheck+0x2d6/0x3c0
  ? submit_bio_noacct+0x47/0x4c0
  submit_bio_wait+0x48/0xa0
  __blkdev_direct_IO_simple+0xee/0x210
  ? current_time+0x1d/0x100
  ? current_time+0x1d/0x100
  ? __bio_clone+0xb0/0xb0
  blkdev_read_iter+0xbb/0x140
  vfs_read+0x239/0x310
  ksys_read+0x58/0xc0
  do_syscall_64+0x6c/0x180
  entry_SYSCALL_64_after_hwframe+0x4b/0x53

Signed-off-by: Keith Busch <kbusch@kernel.org>
---
v2->v3:

  Use a helper function to test if an integrity payload needs to be
  checked. And use all the different kinds of check flags.

  Allocate an unchecked buffer only if the integrity format requires it.
  The nvme formats that don't work with PRACT is the first subscriber to
  this new flag.

 block/bio-integrity-auto.c    | 44 +++++++++++++++++++++++------------
 drivers/nvme/host/core.c      |  3 +++
 include/linux/blk-integrity.h |  1 +
 3 files changed, 33 insertions(+), 15 deletions(-)

diff --git a/block/bio-integrity-auto.c b/block/bio-integrity-auto.c
index e524c609be506..da07212087e06 100644
--- a/block/bio-integrity-auto.c
+++ b/block/bio-integrity-auto.c
@@ -43,6 +43,12 @@ static void bio_integrity_verify_fn(struct work_struct *work)
 	bio_endio(bio);
 }
 
+#define BIP_CHECK_FLAGS (BIP_CHECK_GUARD | BIP_CHECK_REFTAG | BIP_CHECK_APPTAG)
+static bool bip_should_check(struct bio_integrity_payload *bip)
+{
+	return bip->bip_flags & BIP_CHECK_FLAGS;
+}
+
 /**
  * __bio_integrity_endio - Integrity I/O completion function
  * @bio:	Protected bio
@@ -54,12 +60,12 @@ static void bio_integrity_verify_fn(struct work_struct *work)
  */
 bool __bio_integrity_endio(struct bio *bio)
 {
-	struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
 	struct bio_integrity_payload *bip = bio_integrity(bio);
 	struct bio_integrity_data *bid =
 		container_of(bip, struct bio_integrity_data, bip);
 
-	if (bio_op(bio) == REQ_OP_READ && !bio->bi_status && bi->csum_type) {
+	if (bio_op(bio) == REQ_OP_READ && !bio->bi_status &&
+	    bip_should_check(bip)) {
 		INIT_WORK(&bid->work, bio_integrity_verify_fn);
 		queue_work(kintegrityd_wq, &bid->work);
 		return false;
@@ -84,6 +90,7 @@ bool bio_integrity_prep(struct bio *bio)
 {
 	struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
 	struct bio_integrity_data *bid;
+	bool set_flags = true;
 	gfp_t gfp = GFP_NOIO;
 	unsigned int len;
 	void *buf;
@@ -100,19 +107,24 @@ bool bio_integrity_prep(struct bio *bio)
 
 	switch (bio_op(bio)) {
 	case REQ_OP_READ:
-		if (bi->flags & BLK_INTEGRITY_NOVERIFY)
-			return true;
+		if (bi->flags & BLK_INTEGRITY_NOVERIFY) {
+			if (!(bi->flags & BLK_INTEGRITY_BUFFER_REQUIRED))
+				return true;
+			set_flags = false;
+		}
 		break;
 	case REQ_OP_WRITE:
-		if (bi->flags & BLK_INTEGRITY_NOGENERATE)
-			return true;
-
 		/*
 		 * Zero the memory allocated to not leak uninitialized kernel
 		 * memory to disk for non-integrity metadata where nothing else
 		 * initializes the memory.
 		 */
-		if (bi->csum_type == BLK_INTEGRITY_CSUM_NONE)
+		if (bi->flags & BLK_INTEGRITY_NOGENERATE) {
+			if (!(bi->flags & BLK_INTEGRITY_BUFFER_REQUIRED))
+				return true;
+			gfp |= __GFP_ZERO;
+			set_flags = false;
+		} else if (bi->csum_type == BLK_INTEGRITY_CSUM_NONE)
 			gfp |= __GFP_ZERO;
 		break;
 	default:
@@ -137,19 +149,21 @@ bool bio_integrity_prep(struct bio *bio)
 	bid->bip.bip_flags |= BIP_BLOCK_INTEGRITY;
 	bip_set_seed(&bid->bip, bio->bi_iter.bi_sector);
 
-	if (bi->csum_type == BLK_INTEGRITY_CSUM_IP)
-		bid->bip.bip_flags |= BIP_IP_CHECKSUM;
-	if (bi->csum_type)
-		bid->bip.bip_flags |= BIP_CHECK_GUARD;
-	if (bi->flags & BLK_INTEGRITY_REF_TAG)
-		bid->bip.bip_flags |= BIP_CHECK_REFTAG;
+	if (set_flags) {
+		if (bi->csum_type == BLK_INTEGRITY_CSUM_IP)
+			bid->bip.bip_flags |= BIP_IP_CHECKSUM;
+		if (bi->csum_type)
+			bid->bip.bip_flags |= BIP_CHECK_GUARD;
+		if (bi->flags & BLK_INTEGRITY_REF_TAG)
+			bid->bip.bip_flags |= BIP_CHECK_REFTAG;
+	}
 
 	if (bio_integrity_add_page(bio, virt_to_page(buf), len,
 			offset_in_page(buf)) < len)
 		goto err_end_io;
 
 	/* Auto-generate integrity metadata if this is a write */
-	if (bio_data_dir(bio) == WRITE)
+	if (bio_data_dir(bio) == WRITE && bip_should_check(&bid->bip))
 		blk_integrity_generate(bio);
 	else
 		bid->saved_bio_iter = bio->bi_iter;
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index af871d268fcb6..f0992a4e59512 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -1867,6 +1867,9 @@ static bool nvme_init_integrity(struct nvme_ns_head *head,
 		break;
 	}
 
+	if (!nvme_ns_has_pi(head))
+		bi->flags |= BLK_INTEGRITY_BUFFER_REQUIRED;
+
 	bi->tuple_size = head->ms;
 	bi->pi_offset = info->pi_offset;
 	return true;
diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h
index c7eae0bfb013f..ad657d086e715 100644
--- a/include/linux/blk-integrity.h
+++ b/include/linux/blk-integrity.h
@@ -13,6 +13,7 @@ enum blk_integrity_flags {
 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
 	BLK_INTEGRITY_REF_TAG		= 1 << 3,
 	BLK_INTEGRITY_STACKED		= 1 << 4,
+	BLK_INTEGRITY_BUFFER_REQUIRED	= 1 << 5,
 };
 
 const char *blk_integrity_profile_name(struct blk_integrity *bi);
-- 
2.47.1


             reply	other threads:[~2025-05-08 17:58 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-08 17:58 Keith Busch [this message]
2025-05-09  4:19 ` [PATCH] block: always allocate integrity buffer when required Christoph Hellwig
2025-05-09 14:33   ` Keith Busch
2025-05-09 14:39     ` Christoph Hellwig

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20250508175814.1176459-1-kbusch@meta.com \
    --to=kbusch@meta.com \
    --cc=axboe@kernel.dk \
    --cc=hch@lst.de \
    --cc=kbusch@kernel.org \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-nvme@lists.infradead.org \
    --cc=martin.petersen@oracle.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox