[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-xfs
Subject:    Re: [PATCH] vfs: reduce stack usage by shrinking struct kiocb
From:       Zach Brown <zach.brown () oracle ! com>
Date:       2008-04-28 16:13:22
Message-ID: 4815F7A2.7060605 () oracle ! com
[Download RAW message or body]

Denys Vlasenko wrote:
> Hi Al, Benjamin, David,
> 
> struct kiocb is placed on stack by, for example, do_sync_write().
> Eventually it contributes to xfs writeout path's stack usage, among others.
> This is *the* path which causes 4k stack overflows on i386 with xfs.
> 
> This patch trivially reorders fields of this structure,
> and makes some of them smaller.

Great, thanks for doing this.  I see one fatal bug, though.  Can you fix
it up and resubmit?

-	unsigned long		ki_flags;
+	unsigned short		ki_flags; /* range: 0..2 */

Careful, ki_flags is used by the bitops routines which require an
unsigned long.  I'd just leave ki_flags as is.

> ki_nr_segs: ulong -> uint: nobody uses 4 billion element writev's
>                            (and it would not work anyway)

Indeed.  Maybe explicitly mention that it's safe 'cause we pass
ki_nbytes to rw_copy_check_uvector() for comparison against UIO_MAXIOV
before we store it in the kiocb.

+	/*unsigned short	ki_opcode; - moved up for denser packing */

Don't bother commenting out fields that are moved, just move 'em.

Otherwise it looks great, thanks.

- z


[prev in list] [next in list] [prev in thread] [next in thread]