lkml.org 
[lkml]   [2005]   [Mar]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: I/O error propagation
I agree.

But what if the file systems can handle certain errors better than
what the drivers can do now ? Take for e.g., data corruption. If the
driver finds a corrupted sector that it cannot recover, it is going to
convert this specific error in to a more generic error code (-EIO) and
report it to the file system, But the file system might find it is ok
to have a block of data with few corrupted bytes (say, if the block
belongs to a large video file) rather than receive an I/O error code
with no data at all.

What I'm thinking is this: do we have to export richer error codes to
the filesystem from the driver and let the file system handle them as
it thinks appropriate ?

thanks,

On Thu, 3 Mar 2005 15:20:55 -0500 (EST), linux-os <linux-os@analogic.com> wrote:
> On Thu, 3 Mar 2005, V P wrote:
>
> > Hi,
> >
> > I have a question on how disk errors get propagated to
> > the file systems.
> >
> >> From looking at the SCSI/IDE drivers, it looks like there
> > could be many reasons for an I/O to fail. It could be
> > bus timeout, media errors, and so on.
> >
> > Does all these errors get reported to the file system ?
> > It looks like all the different types of errors get
> > turned into a single I/O error (-EIO) and passed on to the
> > file system.
> >
> > Or is there a way where we can export better error codes
> > to the file system ?
> >
> > Any idea/input regarding this is greatly appreciated.
> >
> > Thanks.
>
> It depends upon the disk devices, i.e., IDE SCSI, etc., but in
> general all errors reported by the hardware result in retrying
> the operation. If the retry fails after several (device dependent)
> attempts, the actual error is reported as a kernel message. These
> errors can be retrieved using the `dmesg` command and they
> will usually be retained in some kernel log in /var/log (actual
> log-name is vendor dependent).
>
> Following the Unix convention, any errors reported back upstream,
> eventually to the user, get reported ONLY as something defined
> in /usr/include/errno.h (which includes others, ultimately
> /usr/asm/errno.h).
>
> So, you don't need to reinvent anything. If you have hardware
> errors they will be reported in /var/log/messages (or whatever)
> and if you are making a new driver, you are expected to comply
> with the same protocol.
>
> Cheers,
> Dick Johnson
> Penguin : Linux version 2.6.10 on an i686 machine (5537.79 BogoMips).
> Notice : All mail here is now cached for review by Dictator Bush.
> 98.36% of all statistics are fiction.
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 14:10    [W:0.059 / U:0.404 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site