linux-kernel - Re: [bug] ata subsystem related crash with latest -git

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20071018094128.GP5063@kernel.dk>
Date:	Thu, 18 Oct 2007 11:41:28 +0200
From:	Jens Axboe <jens.axboe@...cle.com>
To:	Jeff Garzik <jeff@...zik.org>
Cc:	Ingo Molnar <mingo@...e.hu>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	linux-kernel@...r.kernel.org, Alan Cox <alan@...rguk.ukuu.org.uk>
Subject: Re: [bug] ata subsystem related crash with latest -git

On Thu, Oct 18 2007, Jeff Garzik wrote:
> Jens Axboe wrote:
>> The sata_mv construct looks a bit odd. Does this work? That last
>
> The sata_mv construct worked just fine before sg chaining :)

Yes I know, but I'm trying to works towards getting rid of sg_last() and
ata_sg_is_last() anyway :-)

>> end_mv_sg test should always be true, just being paranoid...
>> diff --git a/drivers/ata/sata_mv.c b/drivers/ata/sata_mv.c
>> index 4df8311..5397eea 100644
>> --- a/drivers/ata/sata_mv.c
>> +++ b/drivers/ata/sata_mv.c
>> @@ -1138,8 +1138,9 @@ static void mv_fill_sg(struct ata_queued_cmd *qc)
>>  {
>>  	struct mv_port_priv *pp = qc->ap->private_data;
>>  	struct scatterlist *sg;
>> -	struct mv_sg *mv_sg;
>> +	struct mv_sg *mv_sg, *end_mv_sg;
>>  +	end_mv_sg = NULL;
>>  	mv_sg = pp->sg_tbl;
>>  	ata_for_each_sg(sg, qc) {
>>  		dma_addr_t addr = sg_dma_address(sg);
>> @@ -1158,14 +1159,12 @@ static void mv_fill_sg(struct ata_queued_cmd *qc)
>>   			sg_len -= len;
>>  			addr += len;
>> -
>> -			if (!sg_len && ata_sg_is_last(sg, qc))
>> -				mv_sg->flags_size |= cpu_to_le32(EPRD_FLAG_END_OF_TBL);
>> -
>> +			end_mv_sg = mv_sg;
>>  			mv_sg++;
>>  		}
>> -
>>  	}
>> +	if (end_mv_sg)
>> +		end_mv_sg->flags_size |= cpu_to_le32(EPRD_FLAG_END_OF_TBL);
>>  }
>>  
>
> I'm testing a similar patch based on ata_fill_sg()'s method, which 
> basically does something similar to what you've done here (see attached).  
> I had noticed that ata_fill_sg() did not call ata_sg_is_last().
>
> If this fixes the problem, I think the best solution would be to delete 
> ata_sg_is_last().  In the few users that exist, we should be able to 
> eliminate the test programmatically as you and ata_fill_sg() have done -- 
> thereby eliminating a branch per loop in a hotpath.
>
> Off to test the attached...  if that doesn't work I'll try your version, 
> though there shouldn't be much difference.

That should work as well. WRT ata_sg_is_last(), if we go ahead with my
recent sg chaining updates, we can keep the test as it would be a single
conditional and not require any looping.

Let me know when you have tested this!

-- 
Jens Axboe

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/