Skip to contentSkip to conversation

scsi: target: Allow FUA if no write cache enabled

PATCH Superseded Stuart Hayes 2 participants 3 messages Apr 28 – Jun 1

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 scsi: target: Allow FUA if no write cache enabled

Modify target_check_fua() to allow FUA support for devices with no write cache enabled.

Without this patch, accesses with FUA set will be rejected, even though they always go directly to the media when there's no write cache.

This is needed because EDK2 FAT filesystem code sets the FUA bit when writing, regardless of whether the device advertises support of DPOFUA. If a UEFI pre-boot write fails, the filesystem can become inaccessible until a reboot. This can cause linux installs to iSCSI to be unbootable, because some systems have firmware that will try to write a file in a vendor-specific directory under /boot/EFI (such as /boot/EFI/Dell) before attempting to boot to a UEFI boot entry pointing to a file in the FAT filesystem.

---
v2: rebased to latest upstream

 drivers/target/target_core_transport.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/target/target_core_transport.c +2 -1
index fad03a15c969..0e07b224c4d0 100644
@@ -3722,5 +3722,6 @@ target_check_wce(struct se_device *dev)
 bool
 target_check_fua(struct se_device *dev)
 {
-	return target_check_wce(dev) && dev->dev_attrib.emulate_fua_write > 0;
+	/* Allow FUA if there's no write cache */
+	return target_check_wce(dev) ? dev->dev_attrib.emulate_fua_write > 0 : 1;
 }
Signature · 1 line
2.47.3
#1 lore raw

Stuart,

Quoted 3 lines

Without this patch, accesses with FUA set will be rejected, even though they always go directly to the media when there's no write cache.

The spec allows this. However, ...

Quoted 3 lines

This is needed because EDK2 FAT filesystem code sets the FUA bit when writing, regardless of whether the device advertises support of DPOFUA.

^^^ that is clearly a spec violation. What is being done to address this issue?

Also, wrt. the patch itself, please see:

https://sashiko.dev/#/patchset/20260428203938.9738-1-stuart.w.hayes%40gmail.com

Signature · 1 line
Martin K. Petersen
#2 lore raw
Quoted 7 lines · Martin K. Petersen

Stuart,

Quoted 3 lines

Without this patch, accesses with FUA set will be rejected, even though they always go directly to the media when there's no write cache.

Quoted 9 lines

The spec allows this. However, ...

Quoted 3 lines

This is needed because EDK2 FAT filesystem code sets the FUA bit when writing, regardless of whether the device advertises support of DPOFUA.

^^^ that is clearly a spec violation. What is being done to address this issue?

I've gone through SBC and SPC, and I can't find anything that would suggest that this is a spec violation.

And, not that what I think makes any difference, but I don't see the point in requiring an initiator to check DPOFUA before setting FUA on a write that needs to go straight to media... why not just set FUA whenever needed, and if the FUA bit wasn't strictly required to ensure that the write went straight to media, who cares, if the requested behavior is still provided?

Quoted 4 lines

Also, wrt. the patch itself, please see:

https://sashiko.dev/#/patchset/20260428203938.9738-1-stuart.w.hayes%40gmail.com

The AI review suggests moving this fix to sbc_check_dpofua(), so that DPOFUA behavior in the mode page is unchanged, and it just silently allows writes with FUA set even if DPOFUA isn't set.

If there's no objection, I'll submit a new patch that does that. The reason I didn't do that the first time around is that it breaks some of the blktests that specifically check to make sure a command fails if FUA is set when DPOFUA isn't.