Skip to contentSkip to conversation

scsi: storvsc: Support manual scans for all Hyper-V targets

PATCH Applied Laurence Oberman 3 participants 4 messages Jul 23 – Aug 7

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 scsi: storvsc: Support manual scans for all Hyper-V targets

The Fibre Channel transport topology created by storvsc exposes only one dummy remote port per SCSI host. Its scsi_target_id is always zero.

As a result, the FC transport user-scan path looks up the remote port using target ID 0. It cannot initiate a scan for Target 1 or higher. No SCSI command is therefore sent to Hyper-V when userspace explicitly requests a scan of one of these targets.

storvsc itself supports up to STORVSC_FC_MAX_TARGETS and already passes scmnd->device->id to Hyper-V as vm_srb->target_id. Devices on Target 1 and higher work when initially discovered. They can also be rediscovered by a full host scan, such as the scan triggered after an FC port bounce.

Provide a storvsc-specific user_scan callback that uses the exported scsi_scan_target() interface. Iterate over the requested channel and target ranges so that wildcard and explicitly addressed scans retain the expected SCSI sysfs scan semantics.

This bypasses the single-rport lookup in fc_user_scan() and allows userspace to explicitly scan any target supported by storvsc without requiring one synthetic fc_rport for every Hyper-V target.

---

Changes in v2:
- Replace the private, unexported scsi_scan_host_selected() helper with the exported scsi_scan_target() interface.
- Preserve wildcard channel and target scan behavior.
- Verify that storvsc_eh_timed_out() does not use the FC error-handler helpers; the reported missing-rport dereference is not present.

 drivers/scsi/storvsc_drv.c | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)
drivers/scsi/storvsc_drv.c +32 -0
index 571ea549152b..58b6ae96d03f 100644
@@ -520,6 +520,36 @@ static void storvsc_host_scan(struct work_struct *work)
 	scsi_scan_host(host);
 }
 
+#if IS_ENABLED(CONFIG_SCSI_FC_ATTRS)
+static int storvsc_user_scan(struct Scsi_Host *host,
+			     unsigned int channel,
+			     unsigned int id,
+			     u64 lun)
+{
+	unsigned int first_channel, last_channel;
+	unsigned int first_id, end_id;
+	unsigned int ch, target;
+
+	if ((channel != SCAN_WILD_CARD && channel > host->max_channel) ||
+	    (id != SCAN_WILD_CARD && id >= host->max_id) ||
+	    (lun != SCAN_WILD_CARD && lun >= host->max_lun))
+		return -EINVAL;
+
+	first_channel = channel == SCAN_WILD_CARD ? 0 : channel;
+	last_channel = channel == SCAN_WILD_CARD ? host->max_channel : channel;
+	first_id = id == SCAN_WILD_CARD ? 0 : id;
+	end_id = id == SCAN_WILD_CARD ? host->max_id : id + 1;
+
+	for (ch = first_channel; ch <= last_channel; ch++) {
+		for (target = first_id; target < end_id; target++)
+			scsi_scan_target(&host->shost_gendev, ch, target, lun,
+					 SCSI_SCAN_MANUAL);
+	}
+
+	return 0;
+}
+#endif
+
 static void storvsc_remove_lun(struct work_struct *work)
 {
 	struct storvsc_scan_work *wrk;
@@ -2232,6 +2262,8 @@ static int __init storvsc_drv_init(void)
 	fc_transport_template = fc_attach_transport(&fc_transport_functions);
 	if (!fc_transport_template)
 		return -ENODEV;
+
+	fc_transport_template->user_scan = storvsc_user_scan;
 #endif
 
 	ret = vmbus_driver_register(&storvsc_drv);
Signature · 1 line
2.55.0
#1 lore raw
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] Kernel panic in fc_eh_timed_out() due to missing NULL check for fc_rport, contradicting the v2 changelog claim.
- [High] Use-After-Free race condition in fc_user_scan_tgt due to unpinned rport pointer.
--

--- Patch [1]: [PATCH v2] scsi: storvsc: Support manual scans for all Hyper-V targets --- commit 7a84d34942989f4df1a7195141939c24f221c9ea
Author: Laurence Oberman <[email protected]>

scsi: storvsc: Support manual scans for all Hyper-V targets

This patch provides a storvsc-specific user_scan callback that uses the exported scsi_scan_target() interface to allow explicit scanning of targets without requiring synthetic fc_rports for each target.

Quoted 7 lines

---
Changes in v2:
- Replace the private, unexported scsi_scan_host_selected() helper with the exported scsi_scan_target() interface.
- Preserve wildcard channel and target scan behavior.
- Verify that storvsc_eh_timed_out() does not use the FC error-handler helpers; the reported missing-rport dereference is not present.

[Severity: High]
This isn't a bug introduced by this patch, but is the v2 changelog accurate regarding storvsc_eh_timed_out()?

Looking at storvsc_eh_timed_out() in drivers/scsi/storvsc_drv.c, it appears to explicitly route I/O timeouts to the FC transport error handler:

if (scmnd->device->host->transportt == fc_transport_template)

        return fc_eh_timed_out(scmnd);

When fc_eh_timed_out() handles this, it extracts the fc_rport pointer:

struct fc_rport *rport = starget_to_rport(scsi_target(scmd->device)); if (rport->port_state == FC_PORTSTATE_BLOCKED)

Because storvsc targets are created without an intermediate fc_rport parent, starget_to_rport() returns NULL. Does this mean fc_eh_timed_out() will immediately dereference a NULL pointer during an I/O timeout?

As a related observation on the FC transport layer scan paths:

[Severity: High]
This is a pre-existing issue, but does fc_user_scan_tgt() have a race condition that could lead to a use-after-free?

In drivers/scsi/scsi_transport_fc.c:fc_user_scan_tgt(), the host lock is dropped before scanning the target:

spin_unlock_irqrestore(shost->host_lock, flags);
scsi_scan_target(&rport->dev, channel, id, lun,

                 SCSI_SCAN_MANUAL);

If a concurrent FC link down event or port removal triggers fc_rport_final_delete(), can the rport device be removed and freed before or during the scsi_scan_target() execution? If so, should a reference be acquired on rport->dev prior to dropping the lock?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#2 lore raw
Quoted 72 lines · On Thu, 2026-07-23 at 16:51 +0000, [email protected] wrote:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] Kernel panic in fc_eh_timed_out() due to missing NULL check for fc_rport, contradicting the v2 changelog claim.
- [High] Use-After-Free race condition in fc_user_scan_tgt due to unpinned rport pointer.
--

--- Patch [1]: [PATCH v2] scsi: storvsc: Support manual scans for all Hyper-V targets ---
commit 7a84d34942989f4df1a7195141939c24f221c9ea
Author: Laurence Oberman <[email protected]>

scsi: storvsc: Support manual scans for all Hyper-V targets

This patch provides a storvsc-specific user_scan callback that uses the
exported scsi_scan_target() interface to allow explicit scanning of targets
without requiring synthetic fc_rports for each target.

Quoted 9 lines

---
Changes in v2:
- Replace the private, unexported scsi_scan_host_selected() helper with
  the exported scsi_scan_target() interface.
- Preserve wildcard channel and target scan behavior.
- Verify that storvsc_eh_timed_out() does not use the FC error- handler
  helpers; the reported missing-rport dereference is not present.

[Severity: High]
This isn't a bug introduced by this patch, but is the v2 changelog accurate
regarding storvsc_eh_timed_out()?

Looking at storvsc_eh_timed_out() in drivers/scsi/storvsc_drv.c, it appears
to explicitly route I/O timeouts to the FC transport error handler:

if (scmnd->device->host->transportt == fc_transport_template) return fc_eh_timed_out(scmnd);

When fc_eh_timed_out() handles this, it extracts the fc_rport pointer:

struct fc_rport *rport = starget_to_rport(scsi_target(scmd->device)); if (rport->port_state == FC_PORTSTATE_BLOCKED)

Because storvsc targets are created without an intermediate fc_rport parent,
starget_to_rport() returns NULL. Does this mean fc_eh_timed_out() will
immediately dereference a NULL pointer during an I/O timeout?

As a related observation on the FC transport layer scan paths:

[Severity: High]
This is a pre-existing issue, but does fc_user_scan_tgt() have a race condition that could lead to a use-after-free?

In drivers/scsi/scsi_transport_fc.c:fc_user_scan_tgt(), the host lock is
dropped before scanning the target:

spin_unlock_irqrestore(shost->host_lock, flags);
scsi_scan_target(&rport->dev, channel, id, lun,
                 SCSI_SCAN_MANUAL);

Quoted 6 lines

If a concurrent FC link down event or port removal triggers fc_rport_final_delete(), can the rport device be removed and freed before
or during the scsi_scan_target() execution? If so, should a reference be
acquired on rport->dev prior to dropping the lock?

I do not see any of F/C related calls in what I modified or in the code I worked on so for me v2 is the correct and final version of this.

Thanks
Laurence

maintainer Applied
#3 lore raw
In reply to Laurence Oberman

Laurence,

Quoted 8 lines

The Fibre Channel transport topology created by storvsc exposes only one dummy remote port per SCSI host. Its scsi_target_id is always zero.

As a result, the FC transport user-scan path looks up the remote port using target ID 0. It cannot initiate a scan for Target 1 or higher. No SCSI command is therefore sent to Hyper-V when userspace explicitly requests a scan of one of these targets.

Applied to 7.3/scsi-staging, thanks!

Signature · 1 line
Martin K. Petersen