Hi Jérémy! On Tue, 28 Aug 2007 11:50:12 +0200, Jérémy Bobbio wrote: > On Fri, Aug 17, 2007 at 09:36:29PM +0200, Luca Capello wrote: >> 4) again, at reboot, the machine didn't boot, because the previous >> /dev/sda1 is now /dev/sdb1. I discovered the real problem: AFAIK >> the linux kernel erroneously maps the Promise TX4 devices, >> counting from the last one. I can confirm this because I >> attached a third SATA HD to the port3/channel3 and this one was >> marked as /dev/sda instead of the expected /dev/sdc. > > According to a recent mail on LKML [1], this should be fixed in Linux > 2.6.22. [...] > Could you try it and see if the ordering is correct with the newer > kernel? That would be definitely helpful. :) It seems that the unstable 2.6.22-1-486_2.6.22-3 solved this issue, as you can see from the logs attached. BTW, since one of the two disks in the array broke, I used another SATA disk to perform the test, with the advantage that now recognizing the disks is simpler, because they are from different manufacturers. FYI, ST3250620AS is the original HDD in the RAID (thus where etch is installed) plugged on port1/channel1, while HTS541010G9SA00 is the HDD I used for the test, plugged on port2/channel2. IMHO, the correct actions for this bug should be: 1) retitle to "Promise TX4 disk ordering broken" 2) reassign to source:linux-image-2.6 with at least version 6etch1 [1] 3) mark as fixed in version 9 4) add a note to bug #435834 [2] about the fix I'll wait one week from now and then perform the above operations. Thx, bye, Gismo / Luca Footnotes: [1] as I guess this problem is present in all 2.6.18 linux images [2] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=435834
Attachment:
2.6.18_disks.log.gz
Description: Promise TX4 log for 2.6.18
Attachment:
2.6.22_disks.log.gz
Description: Promise TX4 log for 2.6.22