linux-kernel - Re: [PATCH 0/2] Free running cyclic transfer implementation for pl330

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] [day] [month] [year] [list]

Message-Id: <FFF03FEC-4C86-4E2E-BC45-1C8694243203@gmail.com>
Date:   Wed, 26 Apr 2017 20:01:55 +0300
From:   Alexander Kochetkov <al.kochet@...il.com>
To:     Dan Williams <dan.j.williams@...el.com>,
        Vinod Koul <vinod.koul@...el.com>, dmaengine@...r.kernel.org,
        LKML <linux-kernel@...r.kernel.org>,
        linux-rockchip@...ts.infradead.org, hhb <hhb@...k-chips.com>,
        Shawn Lin <shawn.lin@...k-chips.com>,
        王晓腾 <caesar.wang@...k-chips.com>,
        张学广 <sugar.zhang@...k-chips.com>,
        黄涛 <ulysses.huang@...k-chips.com>,
        Stephen Barber <smbarber@...omium.org>,
        Krzysztof Kozlowski <krzk@...nel.org>
Subject: Re: [PATCH 0/2] Free running cyclic transfer implementation for pl330

Hello!

Just to let you know, that I got following test report from Stephen (thanks a lot!).

The patch won’t work due to full cyclic transfer doesn’t fit into mcbufsz (256 bytes long).
His application requested driver to do cyclic transfer with large number of cycles.
pl330 microcode has restriction on how far PC can change and this limit is 256 bytes,
so increasing mcbufsz will not solve problem. Don’t want to make jump bridges or
something like on the fly CPU modified microcode.

> 14 апр. 2017 г., в 23:24, Stephen Barber <smbarber@...omium.org> написал(а):
> 
> Hi Alexander,
> 
> Thanks for your patches!
> 
> I gave them a try on kevin (Samsung Chromebook Plus) with our
> chromeos-4.4 kernel branch, plus some cherry-picks on top to get
> things to apply cleanly. Here's what my tree looks like:
> 
> 344daf13fa4e (HEAD -> apply-pl330) dmaengine: pl330: don't emit code
> for one iteration loop
> 062596b83fec dmaengine: pl330: make cyclic transfer free runnable
> 0e75f2647b8e dmaengine: pl330: fix double lock
> 7d22b54b79f2 dmaengine: pl330: remove unused ‘regs’
> 8769d7115cec dmaengine: pl330: do not generate unaligned access
> 65ad077f685b dmaengine: pl330: convert callback to helper function
> 368e7aa6dffd dmaengine: add support to provide error result from a DMA
> transation
> 8f8afe84472f dmaengine: Add helper function to prep for error reporting
> 2acc1e704232 dmaengine: pl330: explicitly freeup irq
> ed36cde14cf0 (m/master, cros/chromeos-4.4) UPSTREAM: audit: add tty
> field to LOGIN event
> 
> Unfortunately when I start playing audio, things don't seem to work :(
> Rolling back HEAD to "dmaengine: pl330: fix double lock" works though.
> 
> The only thing I get from dmesg relevant to pl330 is this:
> [   59.203375] dma-pl330 ff6d0000.dma-controller:
> pl330_submit_req:1498 Try increasing mcbufsz (12810/256)
> [   59.203395] dma-pl330 ff6d0000.dma-controller: fill_queue:2023 Bad Desc(2)
> [   63.837390] dma-pl330 ff6d0000.dma-controller:
> pl330_submit_req:1498 Try increasing mcbufsz (12810/256)
> [   63.837410] dma-pl330 ff6d0000.dma-controller: fill_queue:2023 Bad Desc(2)
> 
> Thanks,
> Steve
> 


> 14 апр. 2017 г., в 21:04, Krzysztof Kozlowski <krzk@...nel.org> написал(а):
> 
> Let me know if you need more data. I wonder why you haven't experience
> this?

I don’t have idea why this might happen on Odroid and due to the fact the patch don’t
work for Stephen and I don’t have another idea how to implement that, it is better to
leave the problem along.

Krzysztof, thanks a lot for help. Really.

Regards,
Alexander.