Re: kern/47097: ahcisata(4) triggers panic when disk I/O error is caused

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

The following reply was made to PR kern/47097; it has been noted by GNATS.

From: Manuel Bouyer <bouyer%antioche.eu.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: kern-bug-people%NetBSD.org@localhost, gnats-admin%NetBSD.org@localhost, 
netbsd-bugs%NetBSD.org@localhost
Subject: Re: kern/47097: ahcisata(4) triggers panic when disk I/O error is
 caused
Date: Sun, 21 Oct 2012 13:04:20 +0200

 --u3/rZRmxL6MmkK24
 Content-Type: text/plain; charset=us-ascii
 Content-Disposition: inline
 
 On Sat, Oct 20, 2012 at 03:50:00PM +0000, tshiozak%bsdclub.org@localhost wrote:
 > when disk I/O error is caused, ahcisata(4) seems to call tsleep() under 
 > interrupt context and it triggers panic.
 > 
 > call trace is here:
 > 
 > vpanic
 > kern_assert
 > callout_halt
 > sleepq_block
 > ahci_do_reset_drive
 > ahci_reset_drive
 > wddone
 > ahci_bio_complete
 > ahci_intr_port
 > ahci_intr
 
 The attached patch should work around it. But this is going to
 delay() for half a second, which is bad. Something better is needed,
 so we can reset the drive from the atabus thread.
 
 -- 
 Manuel Bouyer <bouyer%antioche.eu.org@localhost>
      NetBSD: 26 ans d'experience feront toujours la difference
 --
 
 --u3/rZRmxL6MmkK24
 Content-Type: text/plain; charset=us-ascii
 Content-Disposition: attachment; filename=diff
 
 Index: ahcisata_core.c
 ===================================================================
 RCS file: /cvsroot/src/sys/dev/ic/ahcisata_core.c,v
 retrieving revision 1.44
 diff -u -p -u -r1.44 ahcisata_core.c
 --- ahcisata_core.c    27 Sep 2012 00:39:47 -0000      1.44
 +++ ahcisata_core.c    21 Oct 2012 11:02:38 -0000
 @@ -721,7 +721,10 @@ again:
                if ((((sig & AHCI_P_TFD_ST) >> AHCI_P_TFD_ST_SHIFT)
                    & WDCS_BSY) == 0)
                        break;
 -              tsleep(&sc, PRIBIO, "ahcid2h", mstohz(10));
 +              if (flags & AT_WAIT)
 +                      tsleep(&sc, PRIBIO, "ahcid2h", mstohz(10));
 +              else
 +                      delay(10000);
        }
        if (i == AHCI_RST_WAIT) {
                aprint_error("%s: BSY never cleared, TD 0x%x\n",
 @@ -740,7 +743,10 @@ again:
            AHCI_READ(sc, AHCI_P_CMD(chp->ch_channel))), DEBUG_PROBE);
  end:
        ahci_channel_stop(sc, chp, flags);
 -      tsleep(&sc, PRIBIO, "ahcirst", mstohz(500));
 +      if (flags & AT_WAIT)
 +              tsleep(&sc, PRIBIO, "ahcirst", mstohz(500));
 +      else
 +              delay(500000);
        /* clear port interrupt register */
        AHCI_WRITE(sc, AHCI_P_IS(chp->ch_channel), 0xffffffff);
        ahci_channel_start(sc, chp, AT_WAIT,
 
 --u3/rZRmxL6MmkK24--