[PATCH as442] gcc 2.96 workaround in sys_getdents64()

[PATCH as442] gcc 2.96 workaround in sys_getdents64()

[Alan Stern]

Andrew:

How serious are people about continuing to support versions of gcc prior 
to 3.0?  My copy of RedHat's gcc-2.96 croaks without this patch:


Signed-off-by: Alan Stern <stern@rowland.harvard.edu>

===== fs/readdir.c 1.26 vs edited =====
--- 1.26/fs/readdir.c	2005-01-09 08:29:46 -05:00
+++ edited/fs/readdir.c	2005-01-13 11:31:58 -05:00
@@ -288,7 +288,7 @@
 	if (lastdirent) {
 		typeof(lastdirent->d_off) d_off = file->f_pos;
 		error = count - buf.count;
-		if (__put_user(d_off, &lastdirent->d_off))
+		if (put_user(d_off, &lastdirent->d_off))
 			error = -EFAULT;
 	}
 

For changes involving code that's not executed very often I wouldn't
hesitate to recommend adopting such a patch.  Does sys_getdents64() fall
in that category?  Also I'm not sure how much extra overhead is added by
calling put_user() rather than __put_user().

So I'll leave it up to you to decide whether the patch should be applied.

Alan Stern

Re: [PATCH as442] gcc 2.96 workaround in sys_getdents64()

[Andreas Dilger]

[-- Attachment #1: Type: text/plain, Size: 1349 bytes --]

On Jan 13, 2005  11:39 -0500, Alan Stern wrote:
> How serious are people about continuing to support versions of gcc prior 
> to 3.0?  My copy of RedHat's gcc-2.96 croaks without this patch:
> 
> 
> Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
> 
> ===== fs/readdir.c 1.26 vs edited =====
> --- 1.26/fs/readdir.c	2005-01-09 08:29:46 -05:00
> +++ edited/fs/readdir.c	2005-01-13 11:31:58 -05:00
> @@ -288,7 +288,7 @@
>  	if (lastdirent) {
>  		typeof(lastdirent->d_off) d_off = file->f_pos;
>  		error = count - buf.count;
> -		if (__put_user(d_off, &lastdirent->d_off))
> +		if (put_user(d_off, &lastdirent->d_off))
>  			error = -EFAULT;
>  	}
>  
> 
> For changes involving code that's not executed very often I wouldn't
> hesitate to recommend adopting such a patch.  Does sys_getdents64() fall
> in that category?  Also I'm not sure how much extra overhead is added by
> calling put_user() rather than __put_user().

The getdents64() syscall is used for every readdir, though I doubt the
extra overhead would be fatal.  In any case, Linus already accepted a
different patch which set "error = -EFAULT" first, then "count - buf.count"
afterward.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://members.shaw.ca/adilger/             http://members.shaw.ca/golinux/


[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]