Re: 6.0-BETA1: ATA RAID rebuild not working

On 25. jul. 2005, at 21.19, Frode Nordahl wrote:

> Hello,
>
> I am having trouble rebuilding my ATA / SATA RAIDs with 6.0-BETA1.
>
> I have tried it on:
>     Intel SCB2 with onboard Promise PDC20267 UDMA100 controller
>     Intel 875WP3 with onboard ICH5 SATA150 controller
>         (Using PseudoRAID, not Intel MatrixRAID...)
>
> # atacontrol status ar0
> ar0: ATA RAID1 subdisks: DOWN ad6 status: DEGRADED
>
> # atacontrol addspare ar0 ad4
> ad4: inserted into ar0 disk0 as spare
>
> # atacontrol rebuild ar0
>
> # atacontrol status ar0
> ar0: ATA RAID1 subdisks: ad4 ad6 status: REBUILDING 0% completed
>
> And then it just sits at 0% forever. I started a rebuild process on  
> the Promise this friday, and it is still at 0%.
>
> During my testing I noticed that when running atacontrol rebuild, a  
> dd process shows up. What is it for?
>
> root    546  0.0  0.3  3240  2760  p0- DNL  Fri01PM   0:00.01 /bin/ 
> dd if=/dev/ar0 of=/dev/null bs=1m
>
> (note, it has run since friday)
>
> Tracing this process show the following:
> sched_switch(c3885300,0,1) at sched_switch+0x177
> mi_switch(1,0) at mi_switch+0x270
> sleepq_switch(d7540f08,ececabc4,c0637469,d7540f08,0) at  
> sleepq_switch+0xe0
> sleepq_wait(d7540f08,0,0,c085c10e,e52) at sleepq_wait+0x30
> bwait(d7540f08,4c,c085429d) at bwait+0x47
> physio(c381be00,ececacbc,0,100000,c381be00) at physio+0x1db
> devfs_read_f(c3859000,ececacbc,c3ae7780,0,c3885300) at devfs_read_f 
> +0x87
> dofileread(c3885300,4,c3859000,ececacbc,ffffffff) at dofileread+0x85
> kern_readv(c3885300,4,ececacbc,804f000,100000) at kern_readv+0x36
> read(c3885300,ececad04,3,1,246) at read+0x45
> syscall(3b,3b,3b,2104,bfbfed40) at syscall+0x22f
> Xint0x80_syscall() at Xint0x80_syscall+0x1f
> --- syscall (3, FreeBSD ELF32, read), eip = 0x280b9727, esp =  
> 0xbfbfec8c, ebp = 0xbfbfecd8 ---
>
> db> show locks
> exclusive sleep mutex Giant r = 0 (0xc091dba0) locked _at_ /usr/src/ 
> sys/kern/kern_intr.c:544
>

I did some detective work to try and narrow down when this stopped  
working:
5.3-RELEASE            OK
5.3-RELEASE w/atamk3n  OK*
5.4-RELEASE            OK
5.4-RELEASE w/atamk3n  NOT OK

*) The dd starts and finishes, but status stays at 0%, and the RAID  
is never marked as clean


So there are two seperate problems here, but I guess it's safe to say  
that something outside the ATA code changed to make rebuild not even  
start in 5.4-RELEASE and beyond.

There seems to be a general problem with rebuild in atamk3n though.

Frode Nordahl
frode_at_nordahl.net