linux-kernel - Re: [PATCH] dmaengine: fsl-edma: fixup reg offset and hw S/G support in big-endian model

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: <45444544.Xl4NeUiTm0@wuerfel>
Date:	Tue, 23 Sep 2014 14:30:30 +0200
From:	Arnd Bergmann <arnd@...db.de>
To:	linux-arm-kernel@...ts.infradead.org
Cc:	Jingchang Lu <jingchang.lu@...escale.com>, vinod.koul@...el.com,
	dmaengine@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] dmaengine: fsl-edma: fixup reg offset and hw S/G support in big-endian model

On Tuesday 23 September 2014 17:15:19 Jingchang Lu wrote:
> @@ -459,20 +480,27 @@ static void fill_tcd_params(struct fsl_edma_engine *edma,
>         u16 csr = 0;
>  
>         /*
> -        * eDMA hardware SGs require the TCD parameters stored in memory
> -        * the same endian as the eDMA module so that they can be loaded
> -        * automatically by the engine
> +        * eDMA hardware SGs requires the TCDs to be auto loaded
> +        * in the same endian as the core whenver the eDAM engine's
> +        * register endian. So we don't swap the value, waitting
> +        * for fsl_set_tcd_params doing the swap.
>          */

This makes no sense: how would the eDMA hardware know what endianess
the CPU is using to write this data? Have you tried this running on
the same hardware with both big-endian and little-endian kernels?

Also, you access this as little-endian data unconditionally rather
than cpu-endian as the comment suggests, maybe that is what you meant
here?

> -       edma_writel(edma, src, &(tcd->saddr));
> -       edma_writel(edma, dst, &(tcd->daddr));
> -       edma_writew(edma, attr, &(tcd->attr));
> -       edma_writew(edma, EDMA_TCD_SOFF_SOFF(soff), &(tcd->soff));
> -       edma_writel(edma, EDMA_TCD_NBYTES_NBYTES(nbytes), &(tcd->nbytes));
> -       edma_writel(edma, EDMA_TCD_SLAST_SLAST(slast), &(tcd->slast));
> -       edma_writew(edma, EDMA_TCD_CITER_CITER(citer), &(tcd->citer));
> -       edma_writew(edma, EDMA_TCD_DOFF_DOFF(doff), &(tcd->doff));
> -       edma_writel(edma, EDMA_TCD_DLAST_SGA_DLAST_SGA(dlast_sga), &(tcd->dlast_sga));
> -       edma_writew(edma, EDMA_TCD_BITER_BITER(biter), &(tcd->biter));
> +       writel(src, &(tcd->saddr));
> +       writel(dst, &(tcd->daddr));
> +
> +       writew(attr, &(tcd->attr));

You seem to have another preexisting bug: The tcd pointer actually does not
point to mmio registers here at all, but instead points to memory that
has been returned from dma_pool_alloc. It is not valid to use writel()
on such memory, that is only meant to work on real MMIO registers.

You should use regular assignments to the registers here, e.g.
	
	tcd->saddr = cpu_to_le32(src);
	tcd->daddr = cpu_to_le32(dst);
	...

	Arnd
--
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/