Skip to contentSkip to conversation

nvme: core: reject invalid LBA data size from Identify Namespace

PATCH Superseded Chao Shi 4 participants 11 messages Apr 18 – May 15 NVMe

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Chao Shi

author
#0 lore raw

PATCH 1/1 nvme: core: reject invalid LBA data size from Identify Namespace

nvme_update_ns_info_block() trusts id->lbaf[lbaf].ds from the controller and assigns it directly to ns->head->lba_shift without bounds checking. nvme_lba_to_sect() then does:

    return lba << (head->lba_shift - SECTOR_SHIFT);

When called with lba = le64_to_cpu(id->nsze) to compute the device capacity, an attacker-controlled controller can choose ds < 9 or a combination of (ds, nsze) that makes the left shift overflow sector_t. The former is a C undefined behaviour that UBSAN reports as a BUG; the latter silently yields a bogus capacity that the block layer then trusts for bounds checking.

Validate ds against SECTOR_SHIFT and use check_shl_overflow() to compute capacity so that any (ds, nsze) combination that would overflow sector_t is rejected. The namespace is skipped with -EIO instead of crashing the kernel. This is reachable by a malicious NVMe device, a buggy firmware, or an attacker-controlled NVMe-oF target.

Stack trace (UBSAN, ds < 9 variant):

  RIP: nvme_lba_to_sect drivers/nvme/host/nvme.h:699 [inline]
  RIP: nvme_update_ns_info_block.cold+0x5/0x7
  Call Trace:
   nvme_update_ns_info+0x175/0xd90 drivers/nvme/host/core.c:2467
   nvme_validate_ns drivers/nvme/host/core.c:4299 [inline]
   nvme_scan_ns drivers/nvme/host/core.c:4350
   nvme_scan_ns_async+0xa5/0xe0 drivers/nvme/host/core.c:4383
   async_run_entry_fn
   process_one_work
   worker_thread
   kthread

Found by Syzkaller.

Fixes 9419e71b8d67 ("nvme: move ns id info to struct nvme_ns_head")
---
 drivers/nvme/host/core.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/nvme/host/core.c +11 -1
index 1e33af94c24..9b3bf3e4075 100644
@@ -2407,9 +2407,19 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	lim = queue_limits_start_update(ns->disk->queue);
 
 	memflags = blk_mq_freeze_queue(ns->disk->queue);
+	if (id->lbaf[lbaf].ds < SECTOR_SHIFT ||
+	    check_shl_overflow(le64_to_cpu(id->nsze),
+			       id->lbaf[lbaf].ds - SECTOR_SHIFT,
+			       &capacity)) {
+		dev_warn_once(ns->ctrl->device,
+			"invalid LBA data size %u, skipping namespace\n",
+			id->lbaf[lbaf].ds);
+		ret = -EIO;
+		blk_mq_unfreeze_queue(ns->disk->queue, memflags);
+		goto out;
+	}
 	ns->head->lba_shift = id->lbaf[lbaf].ds;
 	ns->head->nuse = le64_to_cpu(id->nuse);
-	capacity = nvme_lba_to_sect(ns->head, le64_to_cpu(id->nsze));
 	nvme_set_ctrl_limits(ns->ctrl, &lim, false);
 	nvme_configure_metadata(ns->ctrl, ns->head, id, nvm, info);
 	nvme_set_chunk_sectors(ns, id, &lim);
Signature · 1 line
2.43.0
#1 lore raw
Quoted 1 line · Chao Shi
Fixes 9419e71b8d67 ("nvme: move ns id info to struct nvme_ns_head")

This fixes tag is wrong. The above commit doesn't touch the code line in question.

Probably a5b1cd61820e ("nvme: move a few things out of nvme_update_disk_info") is the better choice here.

#4 lore raw
In reply to Chao Shi
Quoted 50 lines · Chao Shi

nvme_update_ns_info_block() trusts id->lbaf[lbaf].ds from the controller and assigns it directly to ns->head->lba_shift without bounds checking. nvme_lba_to_sect() then does:

    return lba << (head->lba_shift - SECTOR_SHIFT);

When called with lba = le64_to_cpu(id->nsze) to compute the device capacity, an attacker-controlled controller can choose ds < 9 or a combination of (ds, nsze) that makes the left shift overflow sector_t. The former is a C undefined behaviour that UBSAN reports as a BUG; the latter silently yields a bogus capacity that the block layer then trusts for bounds checking.

Validate ds against SECTOR_SHIFT and use check_shl_overflow() to compute capacity so that any (ds, nsze) combination that would overflow sector_t is rejected. The namespace is skipped with -EIO instead of crashing the kernel. This is reachable by a malicious NVMe device, a buggy firmware, or an attacker-controlled NVMe-oF target.

Stack trace (UBSAN, ds < 9 variant):

  RIP: nvme_lba_to_sect drivers/nvme/host/nvme.h:699 [inline]
  RIP: nvme_update_ns_info_block.cold+0x5/0x7
  Call Trace:
   nvme_update_ns_info+0x175/0xd90 drivers/nvme/host/core.c:2467
   nvme_validate_ns drivers/nvme/host/core.c:4299 [inline]
   nvme_scan_ns drivers/nvme/host/core.c:4350
   nvme_scan_ns_async+0xa5/0xe0 drivers/nvme/host/core.c:4383
   async_run_entry_fn
   process_one_work
   worker_thread
   kthread

Found by Syzkaller.

Fixes 9419e71b8d67 ("nvme: move ns id info to struct nvme_ns_head")
---
 drivers/nvme/host/core.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/nvme/host/core.c +0 -0
index 1e33af94c24..9b3bf3e4075 100644

drivers/nvme/host/core.c static int nvme_update_ns_info_block(struct nvme_ns *ns, · line 2416

Quoted 10 lines
drivers/nvme/host/core.c +7 -0
@@ -2407,9 +2407,19 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	lim = queue_limits_start_update(ns->disk->queue);
 
 	memflags = blk_mq_freeze_queue(ns->disk->queue);
+	if (id->lbaf[lbaf].ds < SECTOR_SHIFT ||
+	    check_shl_overflow(le64_to_cpu(id->nsze),
+			       id->lbaf[lbaf].ds - SECTOR_SHIFT,
+			       &capacity)) {
+		dev_warn_once(ns->ctrl->device,
+			"invalid LBA data size %u, skipping namespace\n",
+			id->lbaf[lbaf].ds);

Just a nit:

If I'm reading the NVMe spec correctly, ds == 0 has a special meaning:
'LBA format is not currently available.'
maybe we should use a different dev_warn() for ds == 0 ?

Maurizio

maintainer
#6 lore raw
In reply to Chao Shi

drivers/nvme/host/core.c static int nvme_update_ns_info_block(struct nvme_ns *ns, · line 2417

Quoted 9 lines · Chao Shi
drivers/nvme/host/core.c +8 -0
 	memflags = blk_mq_freeze_queue(ns->disk->queue);
+	if (id->lbaf[lbaf].ds < SECTOR_SHIFT ||
+	    check_shl_overflow(le64_to_cpu(id->nsze),
+			       id->lbaf[lbaf].ds - SECTOR_SHIFT,
+			       &capacity)) {
+		dev_warn_once(ns->ctrl->device,
+			"invalid LBA data size %u, skipping namespace\n",
+			id->lbaf[lbaf].ds);
+		ret = -EIO;

I think ENODEV is more appropriate errno.

drivers/nvme/host/core.c static int nvme_update_ns_info_block(struct nvme_ns *ns, · line 2419

Quoted 2 lines
drivers/nvme/host/core.c +2 -0
+		blk_mq_unfreeze_queue(ns->disk->queue, memflags);
+		goto out;

I don't see any particular reason why we shouldn't validate this value before starting the queue updates and freezing the queue, like we for the ncap field up higher. Doing that would make the error case much simpler. Case in point, you're missing the corresponding queue_limits_cancel_update() for this error case.

Quoted 10 lines
drivers/nvme/host/core.c +1 -1
+	}
 	ns->head->lba_shift = id->lbaf[lbaf].ds;
 	ns->head->nuse = le64_to_cpu(id->nuse);
-	capacity = nvme_lba_to_sect(ns->head, le64_to_cpu(id->nsze));
 	nvme_set_ctrl_limits(ns->ctrl, &lim, false);
 	nvme_configure_metadata(ns->ctrl, ns->head, id, nvm, info);
 	nvme_set_chunk_sectors(ns, id, &lim);
Signature · 1 line
2.43.0
maintainer
#2 lore raw
In reply to Daniel Wagner
Quoted 8 lines · Daniel Wagner
Quoted 1 line · Chao Shi
Fixes 9419e71b8d67 ("nvme: move ns id info to struct nvme_ns_head")

This fixes tag is wrong. The above commit doesn't touch the code line in question.

Probably a5b1cd61820e ("nvme: move a few things out of nvme_update_disk_info") is the better choice here.

We've never validated the 'ds' field though, so not sure anything introduced a regression here that warrants a fixing tag.

author
#7 lore raw

v2 cover nvme: core: reject invalid LBA data size from Identify Namespace

In reply to Chao Shi

Thanks for sharing comments! I did such changes in v2:
- Remove Fixes tag(Keith Busch: ds has never been validated)
- Change -EIO to -ENODEV (Keith Busch)
- Add missing "queue_limits_cancel_update()" in error path (Keith Busch)
- Add warning for ds == 0 (LBA format not available) (Maurizio Lombardi)

---

Quoted 1 line · Keith Busch

I think ENODEV is more appropriate errno.

Fixed in v2.

Quoted 2 lines

you're missing the corresponding queue_limits_cancel_update() for this error case.

Added in both error paths in v2.

Quoted 1 line

This fixes tag is wrong.

Removed.

Quoted 1 line · Daniel Wagner

This fixes tag is wrong. Probably a5b1cd61820e is the better choice here.

Keith pointed out that ds has never been validated so no commit introduced a regression — removed the Fixes tag entirely.

Quoted 1 line · Maurizio Lombardi

ds == 0 has a special meaning: 'LBA format is not currently available.'

Thanks,added dev_warn_once() for ds==0 in v2.

Chao Shi (1):
nvme: core: reject invalid LBA data size from Identify Namespace

 drivers/nvme/host/core.c | 21 ++++++++++++++++++++-
 1 file changed, 20 insertions(+), 1 deletion(-)
Signature · 1 line
2.43.0
author
#8 lore raw

PATCH v2 1/1 nvme: core: reject invalid LBA data size from Identify Namespace

nvme_update_ns_info_block() trusts id->lbaf[lbaf].ds from the controller and assigns it directly to ns->head->lba_shift without bounds checking. nvme_lba_to_sect() then does:

    return lba << (head->lba_shift - SECTOR_SHIFT);

When called with lba = le64_to_cpu(id->nsze) to compute the device capacity, an attacker-controlled controller can choose ds < 9 or a combination of (ds, nsze) that makes the left shift overflow sector_t. The former is a C undefined behaviour that UBSAN reports as a BUG; the latter silently yields a bogus capacity that the block layer then trusts for bounds checking.

Validate ds against SECTOR_SHIFT and use check_shl_overflow() to compute capacity so that any (ds, nsze) combination that would overflow sector_t is rejected. The namespace is skipped with -ENODEV instead of crashing the kernel. This is reachable by a malicious NVMe device, a buggy firmware, or an attacker-controlled NVMe-oF target.

Stack trace (UBSAN, ds < 9 variant):

  RIP: nvme_lba_to_sect drivers/nvme/host/nvme.h:699 [inline]
  RIP: nvme_update_ns_info_block.cold+0x5/0x7
  Call Trace:
   nvme_update_ns_info+0x175/0xd90 drivers/nvme/host/core.c:2467
   nvme_validate_ns drivers/nvme/host/core.c:4299 [inline]
   nvme_scan_ns drivers/nvme/host/core.c:4350
   nvme_scan_ns_async+0xa5/0xe0 drivers/nvme/host/core.c:4383
   async_run_entry_fn
   process_one_work
   worker_thread
   kthread

Found by Syzkaller.

---
 drivers/nvme/host/core.c | 21 ++++++++++++++++++++-
 1 file changed, 20 insertions(+), 1 deletion(-)
drivers/nvme/host/core.c +20 -1
index 1e33af94c24..d1711ef59fb 100644
@@ -2407,9 +2407,28 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	lim = queue_limits_start_update(ns->disk->queue);
 
 	memflags = blk_mq_freeze_queue(ns->disk->queue);
+	if (id->lbaf[lbaf].ds == 0) {
+		dev_warn_once(ns->ctrl->device,
+			"LBA format not available, skipping namespace\n");
+		ret = -ENODEV;
+		queue_limits_cancel_update(ns->disk->queue);
+		blk_mq_unfreeze_queue(ns->disk->queue, memflags);
+		goto out;
+	}
+	if (id->lbaf[lbaf].ds < SECTOR_SHIFT ||
+	    check_shl_overflow(le64_to_cpu(id->nsze),
+			       id->lbaf[lbaf].ds - SECTOR_SHIFT,
+			       &capacity)) {
+		dev_warn_once(ns->ctrl->device,
+			"invalid LBA data size %u, skipping namespace\n",
+			id->lbaf[lbaf].ds);
+		ret = -ENODEV;
+		queue_limits_cancel_update(ns->disk->queue);
+		blk_mq_unfreeze_queue(ns->disk->queue, memflags);
+		goto out;
+	}
 	ns->head->lba_shift = id->lbaf[lbaf].ds;
 	ns->head->nuse = le64_to_cpu(id->nuse);
-	capacity = nvme_lba_to_sect(ns->head, le64_to_cpu(id->nsze));
 	nvme_set_ctrl_limits(ns->ctrl, &lim, false);
 	nvme_configure_metadata(ns->ctrl, ns->head, id, nvm, info);
 	nvme_set_chunk_sectors(ns, id, &lim);
Signature · 1 line
2.43.0
maintainer
#9 lore raw
In reply to Chao Shi
Quoted 8 lines · Chao Shi

On Mon, Apr 20, 2026 at 02:51:00PM -0700, Keith Busch wrote:

Fixed in v2.

Quoted 2 lines

you're missing the corresponding queue_limits_cancel_update() for this error case.

Added in both error paths in v2.

Not quite what I suggested. Instead, if you move up the validation checks, you don't have any cleanup in the error case to do.

And don't bother splitting the error cases. The message about the "0" LBADS is in reference to the the broadcast identification. It's never valid for a controller to return an unusable flbas index value for an attached namespace.

author
#3 lore raw
In reply to Daniel Wagner

Hi Daniel,

Thanks for the catch on the Fixes tag.

Quoted 4 lines

This fixes tag is wrong. The above commit doesn't touch the code line in question.
Probably a5b1cd61820e ("nvme: move a few things out of nvme_update_disk_info") is the better choice here.

Keith subsequently pointed out that ds has never been validated upstream, so no single commit introduced a regression to fix. I dropped the Fixes tag entirely in v2 (and v3 keeps it dropped).

Thanks,
Chao

Quoted 9 lines · Daniel Wagner
Quoted 1 line · Chao Shi
Fixes 9419e71b8d67 ("nvme: move ns id info to struct nvme_ns_head")

This fixes tag is wrong. The above commit doesn't touch the code line in question.

Probably a5b1cd61820e ("nvme: move a few things out of nvme_update_disk_info") is the better choice here.

author
#5 lore raw
In reply to Maurizio Lombardi

Hi Maurizio,

Thanks for the nit on v1.

Quoted 3 lines

If I'm reading the NVMe spec correctly, ds == 0 has a special
meaning: 'LBA format is not currently available.' maybe we should use a different dev_warn() for ds == 0 ?

I added the separate warning in v2, but Keith pointed out on the v2 thread that the "LBA format not available" wording in the spec refers to broadcast identification. For an attached namespace, any unusable flbas index value (including 0) is invalid, so v3 merges both cases into a single check.

Apologies for the back-and-forth.

Also: v2 bounced to your old redhat.com address — using [email protected] from here on.

Thanks,
Chao

Quoted 71 lines · Maurizio Lombardi
Quoted 50 lines · Chao Shi

nvme_update_ns_info_block() trusts id->lbaf[lbaf].ds from the controller and assigns it directly to ns->head->lba_shift without bounds checking. nvme_lba_to_sect() then does:

    return lba << (head->lba_shift - SECTOR_SHIFT);

When called with lba = le64_to_cpu(id->nsze) to compute the device capacity, an attacker-controlled controller can choose ds < 9 or a combination of (ds, nsze) that makes the left shift overflow sector_t. The former is a C undefined behaviour that UBSAN reports as a BUG; the latter silently yields a bogus capacity that the block layer then trusts for bounds checking.

Validate ds against SECTOR_SHIFT and use check_shl_overflow() to compute capacity so that any (ds, nsze) combination that would overflow sector_t is rejected. The namespace is skipped with -EIO instead of crashing the kernel. This is reachable by a malicious NVMe device, a buggy firmware, or an attacker-controlled NVMe-oF target.

Stack trace (UBSAN, ds < 9 variant):

  RIP: nvme_lba_to_sect drivers/nvme/host/nvme.h:699 [inline]
  RIP: nvme_update_ns_info_block.cold+0x5/0x7
  Call Trace:
   nvme_update_ns_info+0x175/0xd90 drivers/nvme/host/core.c:2467
   nvme_validate_ns drivers/nvme/host/core.c:4299 [inline]
   nvme_scan_ns drivers/nvme/host/core.c:4350
   nvme_scan_ns_async+0xa5/0xe0 drivers/nvme/host/core.c:4383
   async_run_entry_fn
   process_one_work
   worker_thread
   kthread

Found by Syzkaller.

Fixes 9419e71b8d67 ("nvme: move ns id info to struct nvme_ns_head")
---
 drivers/nvme/host/core.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/nvme/host/core.c +0 -0
index 1e33af94c24..9b3bf3e4075 100644

drivers/nvme/host/core.c static int nvme_update_ns_info_block(struct nvme_ns *ns, · line 2416

Quoted 10 lines
drivers/nvme/host/core.c +7 -0
@@ -2407,9 +2407,19 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
      lim = queue_limits_start_update(ns->disk->queue);
 
      memflags = blk_mq_freeze_queue(ns->disk->queue);
+     if (id->lbaf[lbaf].ds < SECTOR_SHIFT ||
+         check_shl_overflow(le64_to_cpu(id->nsze),
+                            id->lbaf[lbaf].ds - SECTOR_SHIFT,
+                            &capacity)) {
+             dev_warn_once(ns->ctrl->device,
+                     "invalid LBA data size %u, skipping namespace\n",
+                     id->lbaf[lbaf].ds);

Just a nit:

If I'm reading the NVMe spec correctly, ds == 0 has a special meaning:
'LBA format is not currently available.'
maybe we should use a different dev_warn() for ds == 0 ?

Maurizio

author
#10 lore raw
In reply to Keith Busch

Hi Keith,

Thanks for the follow-up — I misread your v1 comment.

Quoted 2 lines

Not quite what I suggested. Instead, if you move up the validation checks, you don't have any cleanup in the error case to do.

Understood now. In v3 the validation is hoisted above queue_limits_start_update() and blk_mq_freeze_queue(), so the error path is a plain goto out with no cancel_update / unfreeze needed. (Please let me know if I things on the right track)

Quoted 4 lines

And don't bother splitting the error cases. The message about the "0" LBADS is in reference to the broadcast identification. It's never valid for a controller to return an unusable flbas index value for an attached namespace.

Collapsed into a single check in v3.

Sending v3 shortly.

Thanks,
Chao

Quoted 18 lines · Keith Busch
Quoted 8 lines · Chao Shi

On Mon, Apr 20, 2026 at 02:51:00PM -0700, Keith Busch wrote:

Fixed in v2.

Quoted 2 lines

you're missing the corresponding queue_limits_cancel_update() for this error case.

Added in both error paths in v2.

Not quite what I suggested. Instead, if you move up the validation checks, you don't have any cleanup in the error case to do.

And don't bother splitting the error cases. The message about the "0" LBADS is in reference to the the broadcast identification. It's never valid for a controller to return an unusable flbas index value for an attached namespace.