linux-kernel - Re: "Fix ATAPI transfer lengths" causes CD writing regression

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: <20071101141501.3746cec2@the-village.bc.nu>
Date:	Thu, 1 Nov 2007 14:15:01 +0000
From:	Alan Cox <alan@...rguk.ukuu.org.uk>
To:	Jeff Garzik <jeff@...zik.org>
Cc:	Jens Axboe <jens.axboe@...cle.com>, Daniel Drake <dsd@...too.org>,
	linux list <linux-kernel@...r.kernel.org>,
	linux-ide@...r.kernel.org, Tejun Heo <htejun@...il.com>,
	Albert Lee <albertcc@...ibm.com>
Subject: Re: "Fix ATAPI transfer lengths" causes CD writing regression

> We are both half-right.  I reverted my version of that completely, 
> applied your version verbatim, and pushed it upstream.  The result: 
> atapi_xlat _did_ start programming lbam/lbah for DMA (another behavior 
> change), but request-sense DMA path was not changed to program lba[mh] 
> for DMA.

I didn't realise we ever issued reques sense by DMA ?

> > Only if your transfer is actually 8K or more.
> 
> No, that's precisely the problem cause for what Daniel is reporting.  We 
> are setting blimit=xfer_len(==10), when the device wants to return more 
> than 10 bytes.  When set to the old limit (8k), the problem didn't occur.

Lots of *other* problems occur instead. Daniel is reporting that if he
makes a stupid request to a buggy drive he gets a reset and the system
continues happily. Even that reset being a reset not a new command issue
is actually us being excessively paranoid.

Prior to these changes people were reporting things like "I cannot
use my CD-ROM".

So given we have ten years of experience of the method I put in taken
from the old IDE layer versus a good way to make one buggy request work
and break everything else - I *know* which I consider to be the right
path.

And PIO drains of DMA controllers do not work for most controllers, some
will simply hang solid if you touch the data register as part of a DMA
sequence (eg promise 2026x)

Alan
-
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/