ATA regression

[Søren Schmidt]

Ruslan Ermilov wrote:

> I don't have a serial console attached here, so below was cut-n-pasted
> by hands:
> 
> : ata4-master: FAILURE - ATA_IDENTIFY timed out
> : ata4-master: FAILURE - ATA_IDENTIFY timed out
> : ata4-master: FAILURE - ATA_IDENTIFY timed out
> : Trying to mount root from ufs:/dev/ad0a
> : ata4-master: FAILURE - ATA_IDENTIFY timed out
> : Slab at 0xffffff003d7e1f38, freei 15 = 0.
> : panic: Duplicate free of item 0xffffff003d7e1ca8 from zone 0xffffff003ffaf500(g_bio)
> : 
> : cpuid = 0
> : KDB: enter: panic
> : [thread pid 3 tid 100029 ]
> : Stopped at	kdb_enter+0x2f: nop
> : db> where
> : kdb_enter() at ...
> : panic() at ...
> : uma_dbg_free() at ...
> : uma_zfree_arg() at ...
> : g_disk_done() at ...
> : ad_done() at ...
> : ata_completed() at ...
> : g_io_schedule_up() at ...
> : g_up_procbody() at ...
> : fork_exit() at ...
> : fork_trampoline() at ...
> : --- trap 0, rip = 0, rsp = 0xffffffffa509dd00, rbp = 0 ---
> : db>
> 
> A panic can be avoided by reverting the ata-queue.c,v 1.41.
> But even after this, I get ATA_IDENTIFY failures.  And as I
> said, reverting to a somewhat earlier version makes it all
> work.

Hmm, on the exact same controller with two SATA drives on it and stock 
current it "just works" (tm) here, I cannot reproduce the problem...
Are you sure its a stock GENERIC kernel with no local mods/patches ?

-- 

-Søren