kernel-packages team mailing list archive

[Dave Chiluk <1449372@xxxxxxxxxxxxxxxxxx>]

** Description changed:

+ [Impact]
+ 
+  * The move to dev_printk has sufficiently modified the scsi error
+ output that it confuses some automated log monitoring.  This causes some
+ errors to not be definitively connected to a disk.  Context does help,
+ but separate line context is difficult to automatedly parse.
+ 
+ [Test Case]
+ 
+  * Use a misbehaving disk that uses the scsi layer and look in kern.log
+ for what appears to be errantly wrapped lines in the scsi error output
+ 
+ [Regression Potential]
+ 
+  * Minimal both changes are clean cherry-picks.  However as I do not
+ have any misbehaving discs I was only able to boot test this.  I could
+ get fancy, but that would take more time.
+ 
+  * Both patches are upstream starting in utopic and vivid respectively.
+ 
+ [Other Info]
+  
+  * This only partially resolves the issue.  That's because the fix (9e5ed2a5b3) that does solve this is very large, not a clean cherry-pick or easily backportable, and requires too many prerequisites and in my opinion not SRUable.  That fix is available starting in v4.0. My guess is this should be available in wily.
+ 
+ 
+ ----- Original description---- 
+ 
  It seems to me that kernel errors, specifically those from the scsi
  driver, are being wrapped, which can confound automated log monitoring.
  
  Example from dmesg -- the lines I marked with X are being printed on
  separate lines where they previously weren't.
  
  [6461080.699629] sd 0:2:2:0: [sdc] Unhandled sense code
  [6461080.699646] sd 0:2:2:0: [sdc]
  [6461080.699650] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE X
  [6461080.699653] sd 0:2:2:0: [sdc]
  [6461080.699655] Sense Key : Medium Error [current] X
  [6461080.699661] sd 0:2:2:0: [sdc]
  [6461080.699663] Add. Sense: No additional sense information X
  [6461080.699667] sd 0:2:2:0: [sdc] CDB:
  [6461080.699669] Read(16): 88 00 00 00 00 00 0c 23 92 e0 00 00 00 08 00 00 X
  [6461080.699684] end_request: I/O error, dev sdc, sector 203657952
  
  This appears to affect 3.2 to pre-4.0, and was sane in lucid.  I'm
  certain it affects quantal to utopic, and vivid is partially mitigated.

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1449372

Title:
  Scsi kernel messages are being printed on separate lines.

Status in linux package in Ubuntu:
  Confirmed

Bug description:
  [Impact]

   * The move to dev_printk has sufficiently modified the scsi error
  output that it confuses some automated log monitoring.  This causes
  some errors to not be definitively connected to a disk.  Context does
  help, but separate line context is difficult to automatedly parse.

  [Test Case]

   * Use a misbehaving disk that uses the scsi layer and look in
  kern.log for what appears to be errantly wrapped lines in the scsi
  error output

  [Regression Potential]

   * Minimal both changes are clean cherry-picks.  However as I do not
  have any misbehaving discs I was only able to boot test this.  I could
  get fancy, but that would take more time.

   * Both patches are upstream starting in utopic and vivid
  respectively.

  [Other Info]
   
   * This only partially resolves the issue.  That's because the fix (9e5ed2a5b3) that does solve this is very large, not a clean cherry-pick or easily backportable, and requires too many prerequisites and in my opinion not SRUable.  That fix is available starting in v4.0. My guess is this should be available in wily.

  
  ----- Original description---- 

  It seems to me that kernel errors, specifically those from the scsi
  driver, are being wrapped, which can confound automated log
  monitoring.

  Example from dmesg -- the lines I marked with X are being printed on
  separate lines where they previously weren't.

  [6461080.699629] sd 0:2:2:0: [sdc] Unhandled sense code
  [6461080.699646] sd 0:2:2:0: [sdc]
  [6461080.699650] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE X
  [6461080.699653] sd 0:2:2:0: [sdc]
  [6461080.699655] Sense Key : Medium Error [current] X
  [6461080.699661] sd 0:2:2:0: [sdc]
  [6461080.699663] Add. Sense: No additional sense information X
  [6461080.699667] sd 0:2:2:0: [sdc] CDB:
  [6461080.699669] Read(16): 88 00 00 00 00 00 0c 23 92 e0 00 00 00 08 00 00 X
  [6461080.699684] end_request: I/O error, dev sdc, sector 203657952

  This appears to affect 3.2 to pre-4.0, and was sane in lucid.  I'm
  certain it affects quantal to utopic, and vivid is partially
  mitigated.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1449372/+subscriptions