Skip to contentSkip to conversation

dm-crypt: Document new no_workqueue flags.

PATCH In review Milan Broz 2 participants 4 messages Aug 20, 2020 – Aug 24, 2020

Subsystems (2)
lore
Share & tools
lore

Milan Broz

author
#0 lore raw

PATCH 1/1 dm-crypt: Document new no_workqueue flags.

Patch 39d42fa96ba1b7d2544db3f8ed5da8fb0d5cb877 introduced new dm-crypt no_read_workqueue and no_write_workqueue flags.

This patch adds documentation to admin guide for them.

---
 Documentation/admin-guide/device-mapper/dm-crypt.rst | 6 ++++++
 1 file changed, 6 insertions(+)
Documentation/admin-guide/device-mapper/dm-crypt.rst +6 -0
index 8f4a3f889d43..94466921415d 100644
@@ -121,6 +121,12 @@ submit_from_crypt_cpus
     thread because it benefits CFQ to have writes submitted using the
     same context.
 
+no_read_workqueue
+    Bypass dm-crypt internal workqueue and process read requests synchronously.
+
+no_write_workqueue
+    Bypass dm-crypt internal workqueue and process write requests synchronously.
+
 integrity:<bytes>:<type>
     The device requires additional <bytes> metadata per-sector stored
     in per-bio integrity structure. This metadata must by provided
Signature · 1 line
2.28.0
#1 lore raw
Quoted 15 lines · Milan Broz

Patch 39d42fa96ba1b7d2544db3f8ed5da8fb0d5cb877 introduced new dm-crypt no_read_workqueue and no_write_workqueue flags.

This patch adds documentation to admin guide for them.

---
 Documentation/admin-guide/device-mapper/dm-crypt.rst | 6 ++++++
 1 file changed, 6 insertions(+)
Documentation/admin-guide/device-mapper/dm-crypt.rst +0 -0
index 8f4a3f889d43..94466921415d 100644

Documentation/admin-guide/device-mapper/dm-crypt.rst submit_from_crypt_cpus · line 128

Quoted 8 lines
Documentation/admin-guide/device-mapper/dm-crypt.rst +5 -0
@@ -121,6 +121,12 @@ submit_from_crypt_cpus
     thread because it benefits CFQ to have writes submitted using the
     same context.
 
+no_read_workqueue
+    Bypass dm-crypt internal workqueue and process read requests synchronously.
+
+no_write_workqueue
+    Bypass dm-crypt internal workqueue and process write requests synchronously.

Could you add one more line here saying something like:

This option is automatically enabled for host-managed zoned block devices (e.g. host-managed SMR hard-disks).

Thanks !

Quoted 5 lines
Documentation/admin-guide/device-mapper/dm-crypt.rst +1 -0
+
 integrity:<bytes>:<type>
     The device requires additional <bytes> metadata per-sector stored
     in per-bio integrity structure. This metadata must by provided
Signature · 2 lines
Damien Le Moal
Western Digital Research
author
#2 lore raw
Quoted 27 lines · Damien Le Moal
Quoted 15 lines · Milan Broz

Patch 39d42fa96ba1b7d2544db3f8ed5da8fb0d5cb877 introduced new dm-crypt no_read_workqueue and no_write_workqueue flags.

This patch adds documentation to admin guide for them.

---
 Documentation/admin-guide/device-mapper/dm-crypt.rst | 6 ++++++
 1 file changed, 6 insertions(+)
Documentation/admin-guide/device-mapper/dm-crypt.rst +0 -0
index 8f4a3f889d43..94466921415d 100644

Documentation/admin-guide/device-mapper/dm-crypt.rst submit_from_crypt_cpus · line 128

Quoted 8 lines
Documentation/admin-guide/device-mapper/dm-crypt.rst +5 -0
@@ -121,6 +121,12 @@ submit_from_crypt_cpus
     thread because it benefits CFQ to have writes submitted using the
     same context.
 
+no_read_workqueue
+    Bypass dm-crypt internal workqueue and process read requests synchronously.
+
+no_write_workqueue
+    Bypass dm-crypt internal workqueue and process write requests synchronously.

Could you add one more line here saying something like:

Quoted 2 lines

This option is automatically enabled for host-managed zoned block devices (e.g. host-managed SMR hard-disks).

Yes, Mike can squeeze it there (Mike, if you prefer, I can resend the patch with the note above).

I just see one problem here - when we activate device without these flags, then dm-crypt decides to set the feature bits because of zone device.

So you activate device with some feature set but table will show something different. It is not a technical problem, but I think DM table was not expected to behave this way.

It is probably not the first change (I think some flags are activated in dm-integrity according to on-disk superblock state only).

Milan

p.s. Both noqueue flags are now supported in cryptsetup master, for LUKS2 we can store them persistently (IOW to be used in all later LUKS2 activations if kernel supports it).
It should appear in next stable cryptsetup release.

#3 lore raw
Quoted 31 lines · Milan Broz
Quoted 18 lines · Damien Le Moal
Quoted 17 lines · Milan Broz

Patch 39d42fa96ba1b7d2544db3f8ed5da8fb0d5cb877 introduced new dm-crypt no_read_workqueue and no_write_workqueue flags.

This patch adds documentation to admin guide for them.

Documentation/admin-guide/device-mapper/dm-crypt.rst | 6 ++++++ 1 file changed, 6 insertions(+)

(quoted diff) +0 -0

b/Documentation/admin-guide/device-mapper/dm-crypt.rst index 8f4a3f889d43..94466921415d 100644 --- a/Documentation/admin-guide/device-mapper/dm-crypt.rst +++ b/Documentation/admin-guide/device-mapper/dm-crypt.rst @@ -121,6 +121,12 @@ submit_from_crypt_cpus thread because it benefits CFQ to have writes submitted using the same context.

Quoted 8 lines
Quoted 3 lines

+no_read_workqueue + Bypass dm-crypt internal workqueue and process
read requests synchronously. + +no_write_workqueue + Bypass dm-crypt
internal workqueue and process write requests synchronously.

Could you add one more line here saying something like:

This option is automatically enabled for host-managed zoned block devices (e.g. host-managed SMR hard-disks).

Yes, Mike can squeeze it there (Mike, if you prefer, I can resend the patch with the note above).

Quoted 6 lines

I just see one problem here - when we activate device without these flags, then dm-crypt decides to set the feature bits because of zone device.

So you activate device with some feature set but table will show something different. It is not a technical problem, but I think DM table was not expected to behave this way.

Hmmm. Never wondered about that... Mike ? Is this a big problem ?

Quoted 8 lines

It is probably not the first change (I think some flags are activated in dm-integrity according to on-disk superblock state only).

Milan

p.s. Both noqueue flags are now supported in cryptsetup master, for LUKS2 we can store them persistently (IOW to be used in all later LUKS2 activations if kernel supports it). It should appear in next stable cryptsetup release.

Great. Thanks. Note that I have code for Luks format on host-managed SMR disks, or rather any host-managed zoned block device, including ZNS NVMe SSDs.

The current cryptsetup code needs some massaging for these drives because:
1) offset+size needs to be zoned aligned, so at the very least defaulting to --offset=<device zone size> is needed.
2) The on-disk formatting is annoyingly not writing sequentially :) So the current code fails if the drive does not have a conventional zone at LBA 0.

My code is rather dirty for now and needs a good cleanup. I will do that rebasing on the next stable that adds the no queue flags support. The current cryptsetup does work as is with plain type, and with Luks type for SMR drives that have a conventional zone at LBA 0 with the added --offset=<device zone size> option specified.

Best regards.

Signature · 2 lines
Damien Le Moal
Western Digital Research