lkml.org 
[lkml]   [1998]   [Jul]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: 2.1.111: IDE DMA disabled...BLAH...BLAH...
    On Wed, 29 Jul 1998, Alan Cox wrote:

    > Linus - Yes there is something wrong
    > No it cannot be DMA if DMA goes back to 1.3.20, almost every report
    > says it started in 2.1.x, most are not UDMA but older DMA
    >
    > The only common theme right now seems to be SMP and SCSI, and thats SMP
    > machines in the SCSI case (not option), unknown in the IDE case.

    My system: 2 x DPT2044UW, 1 x Future Domain PCI, 2 IDE drives each
    on a different bus. Have had crashes seemingly due to memory corruption.
    Once root IDE drive (IBM ~6.4GB) was corrupted, twice the first oops
    was in find_buffer (quickly followed by _lots_ more oopses), once
    in tcp_close.

    Initially using Balles* EATA driver crashes would happen under
    load (ckraid operations, copying large trees) but only intermittently.
    Using eata_dma driver _seems_ fine under load. Since then crashes
    have seemed related to load on the Future Domain - but not simple
    load (such as streaming through the Exabyte), it seems there has
    to be significant other SCSI activity. 20-30GB backup/restore and
    probably one crash will have happened.

    Was using 2.1.107 SMP (this is a single PII-400, BX, no IO-APIC).
    Today was using 2.1.107 *UP* with IDE DMA on both drives *disabled*
    with hdparm when it happened. Tonight will switch the Future Domain
    for another DPT.

    1. Anybody using a Future Domain successfully? (later crashes may
    be _solely_ down to the FD)

    2. Anybody using the Balles* EATA driver successfully? (initial
    crashes happened before I added the FD)

    3. Anybody using multiple SCSI adapters under simultaneous load
    successfully? (All crashes seem related to simultaneous load on
    multiple adapters)

    It is possible that both the Balles* EATA driver and the Future
    Domain driver have race conditions which break them eventually.
    However, my current feeling is that the race is higher level
    and loading four or five devices on two or three controllers
    just makes it more likely.

    Why does eata_dma work? Maybe its timing is different enough to
    avoid the race - so far? Maybe its the only driver of the three
    that isn't broken?

    I'm not seeing IDE problems (but then I'm not stressing the IDE
    at all). I have merrily swapped cables, memory, drives etc. to
    eliminate as many of the atoms as possible.

    Mike

    P.S. No, I can't remember how to Balles*. Sorry :-).

    --
    .----------------------------------------------------------------------.
    | Mike Jagdis | Internet: mailto:mike@roan.co.uk |
    | Roan Technology Ltd. | |
    | 54A Peach Street, Wokingham | Telephone: +44 118 989 0403 |
    | RG40 1XG, ENGLAND | Fax: +44 118 989 1195 |
    `----------------------------------------------------------------------'


    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.rutgers.edu
    Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html

    \
     
     \ /
      Last update: 2005-03-22 13:43    [W:0.022 / U:60.448 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site