ATAng atacontrol issue

From: Andre Guibert de Bruet (andy_at_siliconlandmark.com)
Date: 08/31/03

  • Next message: Peter Kostouros: "Inappropriate ioctl for device (CDIOCREADAUDIO)" 
    Date: Sat, 30 Aug 2003 21:56:42 -0400 (EDT)
To: current@freebsd.org

    As reported to sos@, with a little extra debugging at the bottom...

    -----------

    I was playing with atacontrol reinit'ing all of the channels that are
    installed on my machine, when I accidentally tried to reinit channel '4',
    which clearly doesn't exist. Here's what showed up at the serial console:

    ata0: resetting devices ..
    done
    ata1: resetting devices ..
    done
    ata2: resetting devices ..
    done
    ata3: resetting devices ..
    done
    panic: lock (sleep mutex) Giant not locked @ kern/sys_generic.c:678
    cpuid = 1; lapic.id = 01000000
    Debugger("panic")
    Stopped at Debugger+0x4e: xchgl %ebx,in_Debugger.0
    db> tr
    Debugger(c04088ac,1000000,c040bc40,f0e0abe4,100) at Debugger+0x4e
    panic(c040bc40,c041c423,c04080be,c040bfe4,2a6) at panic+0x151
    witness_unlock(c0494c40,8,c040bfe4,2a6,6) at witness_unlock+0x213
    _mtx_unlock_flags(c0494c40,0,c040bfdb,2a6,ca8f9130) at
    _mtx_unlock_flags+0x80
    ioctl(ca8f9130,f0e0ad14,c042237d,3eb,3) at ioctl+0x4f8
    syscall(2f,2f,2f,805e92f,bfbffd36) at syscall+0x253
    Xint0x80_syscall() at Xint0x80_syscall+0x1d
    --- syscall (54, FreeBSD ELF32, ioctl), eip = 0x804995b, esp = 0xbfbff76c, ebp = 0xbfbffc24 ---

    My ata setup has the following devicesL
    acd0: DVDROM <DVD-ROM BDV316C> at ata0-master PIO4
    ad2: 238475MB <WDC WD2500JB-34EVA0> [484521/16/63] at ata1-master UDMA100
    ad4: 194481MB <Maxtor 6Y200P0> [395136/16/63] at ata2-master UDMA133
    ad6: 194481MB <Maxtor 6Y200P0> [395136/16/63] at ata3-master UDMA133

    These devices are located on the following controllers:

    atapci0: <AMD 768 UDMA100 controller> port 0xb800-0xb80f at device 7.1 on pci0
    ata0: at 0x1f0 irq 14 on atapci0
    ata1: at 0x170 irq 15 on atapci0
    atapci1: <Promise PDC20269 UDMA133 controller> port 0x9000-0x900f,0x9400-0x9403,0x9800-0x9807,0xa000-0xa003,0xa400-0xa407 mem 0xe6000000-0xe6003fff irq 11 at device 5.0 on pci2
    ata2: at 0xa400 on atapci1
    ata3: at 0x9800 on atapci1

    A little bit of GDB'ing gives the following:

    (kgdb) l *(ioctl+0x4f8)
    0xc0270fa8 is in ioctl (../../../kern/sys_generic.c:679).
    674 if (memp)
    675 free(memp, M_IOCTLOPS);
    676 fdrop(fp, td);
    677 done:
    678 mtx_unlock(&Giant);
    679 return (error);
    680 }
    681
    682 /*
    683 * sellock and selwait are initialized in selectinit() via SYSINIT.
    (kgdb) l *(syscall+0x253)
    0xc03c6103 is in syscall (../../../i386/i386/trap.c:1005).
    1000 td->td_retval[0] = 0;
    1001 td->td_retval[1] = frame.tf_edx;
    1002
    1003 STOPEVENT(p, S_SCE, narg);
    1004
    1005 error = (*callp->sy_call)(td, args);
    1006 }
    1007
    1008 switch (error) {
    1009 case 0:
    (kgdb)

    We're not locking Giant in this case because of the following check:

            /*
             * Try to run the syscall without Giant if the syscall
             * is MP safe.
             */
            if ((callp->sy_narg & SYF_MPSAFE) == 0)
                    mtx_lock(&Giant);

    It appears that the syscall is marked as MPSAFE but ends up calling code
    in ioctl that expects Giant to be locked. Any ideas?

    Regards,

    > Andre Guibert de Bruet | Enterprise Software Consultant >
    > Silicon Landmark, LLC. | http://siliconlandmark.com/ >
    _______________________________________________
    freebsd-current@freebsd.org mailing list
    http://lists.freebsd.org/mailman/listinfo/freebsd-current
    To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"

  • Next message: Peter Kostouros: "Inappropriate ioctl for device (CDIOCREADAUDIO)"