Re: [PATCH v21 04/13] xfs: Handle krealloc errors in xlog_recover_add_to_cont_trans

["Darrick J. Wong" <djwong@kernel.org>]

On Wed, Jul 07, 2021 at 03:21:02PM -0700, Allison Henderson wrote:
> Because xattrs can be over a page in size, we need to handle possible
> krealloc errors to avoid warnings.  If the allocation does fail, fall
> back to kmem_alloc_large, with a memcpy.
> 
> The warning:
>    WARNING: CPU: 1 PID: 20255 at mm/page_alloc.c:3446
>                  get_page_from_freelist+0x100b/0x1690
> 
> is caused when sizes larger that a page are allocated with the
> __GFP_NOFAIL flag option.  We encounter this error now because attr
> values can be up to 64k in size.  So we cannot use __GFP_NOFAIL, and
> we need to handle the error code if the allocation fails.
> 
> Signed-off-by: Allison Henderson <allison.henderson@oracle.com>

I'm pretty sure that 'mm: Add kvrealloc' fixes this a little more
elegantly, but either look fine to me.  So while I'll probably take
Dave's, here's a:

Reviewed-by: Darrick J. Wong <djwong@kernel.org>

in the meantime.

--D

> ---
>  fs/xfs/xfs_log_recover.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
> index ec4ccae..6ab467b 100644
> --- a/fs/xfs/xfs_log_recover.c
> +++ b/fs/xfs/xfs_log_recover.c
> @@ -2062,7 +2062,15 @@ xlog_recover_add_to_cont_trans(
>  	old_ptr = item->ri_buf[item->ri_cnt-1].i_addr;
>  	old_len = item->ri_buf[item->ri_cnt-1].i_len;
>  
> -	ptr = krealloc(old_ptr, len + old_len, GFP_KERNEL | __GFP_NOFAIL);
> +	ptr = krealloc(old_ptr, len + old_len, GFP_KERNEL);
> +	if (ptr == NULL) {
> +		ptr = kmem_alloc_large(len + old_len, KM_ZERO);
> +		if (ptr == NULL)
> +			return -ENOMEM;
> +
> +		memcpy(ptr, old_ptr, old_len);
> +	}
> +
>  	memcpy(&ptr[old_len], dp, len);
>  	item->ri_buf[item->ri_cnt-1].i_len += len;
>  	item->ri_buf[item->ri_cnt-1].i_addr = ptr;
> -- 
> 2.7.4
>