Re: [PATCH] cleanup for fixing get_super() races

From: Alexander Viro (
Date: Sat Apr 28 2001 - 10:00:46 EST

On Sat, 28 Apr 2001, Martin Dalecki wrote:

> I think in the context you are inventig the proposed function,
> the drivers has allways an inode at hand. And contrary to what Linus

Read the patch. Almost all cases are of the "loop over partitions of foo"

> says, drivers not just know about the devices they handle, they
> know about the data they should get - at least in the context
> of block devices. And then you could as well pass the inode, which
> is already containing a refference to the corresponding sb and
> save the whole get_super linear array lookup 8-). I think

No, you don't. Moreover, inode of device (even if you had it) _doesn't_
contain a reference to sb of filesystem mounted from that device.

> the less kdev_t the better! It's overused already anyway, like
> for example in the whole SCSI code, where the functions in reality only
> want to pass the minor number to differentiate they behaviour...
> If you are gogin to flag the behaviour of the function,
> then please use a bitpattern of well definded flags as a parameter,
> in a similiar way like it's done for example in many GUI libraries
> (GTK, Motif and so on). This would make it far more readabel.

/me looks at From:
OK, Albert, what have you done with real Martin?

OK, whoever you are - no, "expandable" interfaces of that sort are
rotten idea. What we really need is to replace sync_dev with fsync_dev -
it _is_ correct in such context. That's it - 1 bit of information, no
bitmaps needed.

/me is still boggled by the idea of somebody refereing to GTK as an
example of style...

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

This archive was generated by hypermail 2b29 : Mon Apr 30 2001 - 21:00:20 EST