Re: [patch-2.3.32-pre3] lock_kernel() in fsync(2)
Manfred (manfreds@colorfullife.com)
Mon, 13 Dec 1999 21:24:23 +0100
From: Tigran Aivazian <tigran@sco.COM>
> Both fget() and fput() seem to be reentrant so even if we assume that
> down(&inode->i_sem) is not enough for fop->fsync(), we could
> move lock/unlock_kernel() closer to it like this, right (fs/buffer.c)?
> /* We need to protect against concurrent writers.. */
> + lock_kernel();
> down(&inode->i_sem);
> err = file->f_op->fsync(file, dentry);
> up(&inode->i_sem);
> + unlock_kernel();
>
I'd first get the per-file semaphore, then the big kernel lock: you cannot
dead-lock, and the big kernel lock is more important.
But if you start to fix the lock_kernel calls, could you also fix
sys_lseek() [same fix as above], sys_pread(), sys_pwrite() [can run without
the kernel lock], sys_readv()/sys_writev() [everything except
sock_readv_writev() can run without the kernel lock], old_mmap [in
arch/i386/kernel/sys_i386.c, just reorder fget() and lock_kernel()]
;)
Manfred
-
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.tux.org/lkml/