Re: Mount error on 32 bits, ok on 64 bits

[Jean-Denis Girard <jd.girard@sysnux.pf>]

Le 02/10/2017 à 11:46, Goffredo Baroncelli a écrit :
> On 10/02/2017 07:59 PM, Jean-Denis Girard wrote:
>> Le 30/09/2017 à 17:29, Jean-Denis Girard a écrit :
>>> Le 28/09/2017 à 19:26, Jean-Denis Girard a écrit :
>>> The problem seems to come from commit c821e7f3 "pass bytes to
>>> btrfs_bio_alloc" (https://patchwork.kernel.org/patch/9763081/): the
>>> system is now running fine on 4.13.4 with only that patch reverted.
>>
>> Same situation with 4.14-rc3: my system cannot mount root file-system.
>> If I revert the patch, the system boots normally. This is 100%
>> reproducible. How can I help resolve that issue?
> 
> 
> Looking at the patch, it seems suspect this chunk:
> 
> @@ -2798,7 +2798,7 @@  static int submit_extent_page(int op, int op_flags, struct extent_io_tree *tree,
>  		}
>  	}
>  
> -	bio = btrfs_bio_alloc(bdev, sector);
> +	bio = btrfs_bio_alloc(bdev, sector << 9);
>  	bio_add_page(bio, page, page_size, offset);
>  	bio->bi_end_io = end_io_func;
>  	bio->bi_private = tree;
> 
> Now sector, is defined as
> 
> 	sector_t   [1]
> 
> which in turn it might be defined as 
> 
> 	unsigned long   [2]
> 
> which on 32bit is 32 bit if CONFIG_LBDAF is _not_ defined (CONFIG_LBDAF == Support for large (2TB+) block devices and files)
> 
> The point is that 
> 
> 	sector << 9 
> 
> may overflow if the disk is bigger than 4GB (and in your case it seems to be 8 GB).
> 
> If I am correct, could you please so kindly to 
> 
> - repllay the patch
> - AND try to replace
> 	bio = btrfs_bio_alloc(bdev, sector << 9);
> with
> 	bio = btrfs_bio_alloc(bdev, (u64)sector << 9);

Hi Goffredo,

Thanks for your reply, analysis and correction: it does work, tested on
4.13-rc3, my system boots fine.


Best regards,
-- 
Jean-Denis Girard

SysNux                   Systèmes   Linux   en   Polynésie  française
https://www.sysnux.pf/   Tél: +689 40.50.10.40 / GSM: +689 87.797.527