linux-kernel - Re: [PATCH 06/13] DMAENGINE: driver for the ARM PL080/PL081 PrimeCells

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]

Date:	Tue, 21 Dec 2010 22:25:32 +0000
From:	Russell King - ARM Linux <linux@....linux.org.uk>
To:	Linus Walleij <linus.walleij@...ricsson.com>
Cc:	Viresh Kumar <viresh.kumar@...com>,
	Kukjin Kim <kgene.kim@...sung.com>, yuanyabin1978@...a.com,
	linux-kernel@...r.kernel.org, Ben Dooks <ben-linux@...ff.org>,
	Peter Pearse <peter.pearse@....com>,
	Dan Williams <dan.j.williams@...el.com>,
	linux-arm-kernel@...ts.infradead.org,
	Alessandro Rubini <rubini@...pv.it>
Subject: Re: [PATCH 06/13] DMAENGINE: driver for the ARM PL080/PL081
	PrimeCells

On Tue, Dec 21, 2010 at 06:20:37PM +0000, Russell King - ARM Linux wrote:
> Having just looked at this while trying to undo the DMA API abuses
> in the PL011 UART driver, I'm getting rather frustrated with this
> code.

Here's another couple of problems:

PL011 driver:

+	dmatx->cookie = desc->tx_submit(desc);
+	if (dma_submit_error(dmatx->cookie)) {
+		/* "Complete" DMA (errorpath) */
+		complete(&dmatx->complete);
+		chan->device->device_control(chan, DMA_TERMINATE_ALL, 0);
+		return dmatx->cookie;
+	}

dmatx->cookie is of type dma_cookie_t, which is a s32.

MMCI driver:

+	dma_cookie_t cookie;
...
+	cookie = desc->tx_submit(desc);
+
+	/* Here overloaded DMA controllers may fail */
+	if (dma_submit_error(cookie))
+		goto unmap_exit;

#define dma_submit_error(cookie) ((cookie) < 0 ? 1 : 0)

So, DMA errors are negative values returned from the tx_submit function.

PL08x tx_submit method:

static dma_cookie_t pl08x_tx_submit(struct dma_async_tx_descriptor *tx)
{
        struct pl08x_dma_chan *plchan = to_pl08x_chan(tx->chan);

        atomic_inc(&plchan->last_issued);
        tx->cookie = atomic_read(&plchan->last_issued);
        /* This unlock follows the lock in the prep() function */
        spin_unlock_irqrestore(&plchan->lock, plchan->lockflags);

        return tx->cookie;
}

So, after 2^31 DMA transactions, the DMA engine layer continues happily
along with the queued DMA operation, but the driver gets a negative
cookie returned, and so bails out.

I'm also left wondering about that atomic_t.  It's incremented and
read in a locked region, which will in itself prevent two concurrent
increments.  The only other places its used is from dma_tx_status()
where it is only ever read, and on the initialization path.  That to
me makes no sense.

I'll see if I can get to looking at the PL08x stuff once I've finished
sorting out the PL011 UART driver DMA support into a state where I'm
happy that it's doing things correctly.
--
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/