Bug 154434

Summary: kernel dm-multipath: Make SCSI sense data available in bio end_io path for dm-emc
Product: [Fedora] Fedora Reporter: Alasdair Kergon <agk>
Component: device-mapper-multipathAssignee: Alasdair Kergon <agk>
Status: CLOSED WONTFIX QA Contact:
Severity: medium Docs Contact:
Priority: medium    
Version: rawhideCC: agk, christophe.varoqui, dmo, dwysocha, lmb, mbroz, michael.hagmann, terescik_john, tranlan
Target Milestone: ---Keywords: FutureFeature
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2010-04-27 13:51:51 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
Bio end_io patch
none
hardware-handler sense data patch
none
Make sense buffer available in SCSI request struct [already upstream]
none
Updated bio-end_io-sense patch none

Description Alasdair Kergon 2005-04-11 17:36:15 UTC 
The attached two patches enable SCSI sense data to be available in the
bio-end_io path and use it in dm-mpath's hwhandler as well as dm-emc.c.

This allows the EMC module to take different action according to what sort of
I/O failure has occurred.

[From Jens Axboe & Lars Marowsky-Bree, Suse]
Comment 1 Alasdair Kergon 2005-04-11 17:36:16 UTC 
Created attachment 112959 [details]
Bio end_io patch
Comment 3 Alasdair Kergon 2005-04-11 17:39:16 UTC 
Created attachment 112961 [details]
hardware-handler sense data patch
Comment 4 Lars Marowsky-Bree 2005-04-20 09:01:35 UTC 
Created attachment 113395 [details]
Make sense buffer available in SCSI request struct [already upstream]

Pre-requisite to having the sense data available in the bio
Comment 5 Lars Marowsky-Bree 2005-04-20 09:04:17 UTC 
Created attachment 113396 [details]
Updated bio-end_io-sense patch

This obsoletes the first patch, but I can't set that flag ;)

s->valid didn't imply what Jens thought, and the struct request_sense is
generally not the right one to use here. This corrects parsing the sense buffer
and setting it in the bio, also sets bi_error correctly if no sense data is
available (by setting it to the errno).
Comment 6 Lars Marowsky-Bree 2005-04-20 09:06:55 UTC 
The hardware handler patch is still correct.
Comment 7 Alasdair Kergon 2005-04-21 17:41:31 UTC 
2 patches added to 2.6.12-rc2-udm1
Comment 8 Alasdair Kergon 2010-04-27 13:51:51 UTC 
Superceded by request-based multipath.