linux-kernel - Re: [PATCH] crypto: caam - do not reset pointer size if caam_ptr

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:   Mon, 25 Nov 2019 07:49:55 +0000
From:   Horia Geanta <horia.geanta@....com>
To:     Iuliana Prodan <iuliana.prodan@....com>,
        Herbert Xu <herbert@...dor.apana.org.au>,
        Aymen Sghaier <aymen.sghaier@....com>
CC:     "David S. Miller" <davem@...emloft.net>,
        "linux-crypto@...r.kernel.org" <linux-crypto@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        dl-linux-imx <linux-imx@....com>,
        Andrey Smirnov <andrew.smirnov@...il.com>,
        Alison Wang <alison.wang@....com>
Subject: Re: [PATCH] crypto: caam - do not reset pointer size if caam_ptr_size
 is 64 bits

On 11/25/2019 12:33 AM, Iuliana Prodan wrote:
> In commit 'a1cf573ee95 ("crypto: caam - select DMA address
> size at runtime")' CAAM pointer size (caam_ptr_size) is changed
> from sizeof(dma_addr_t) to runtime value computed from MCFGR register.
> At some point, the bit for Pointer Size should be reset to 0,
> but only for the case when the CAAM pointer size if 32 bits.
> Therefore, use caam_ptr_size instead of sizeof(dma_addr_t).
> 
The logic is circular, see below.

> Fixes: a1cf573ee95 ("crypto: caam - select DMA address size at runtime")
Please Cc the author(s) when adding Fixes tag.

> Cc: <stable@...r.kernel.org>
> Signed-off-by: Iuliana Prodan <iuliana.prodan@....com>
> ---
>  drivers/crypto/caam/ctrl.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/crypto/caam/ctrl.c b/drivers/crypto/caam/ctrl.c
> index d7c3c38..786eef6 100644
> --- a/drivers/crypto/caam/ctrl.c
> +++ b/drivers/crypto/caam/ctrl.c
> @@ -674,7 +674,7 @@ static int caam_probe(struct platform_device *pdev)
>  		clrsetbits_32(&ctrl->mcr, MCFGR_AWCACHE_MASK | MCFGR_LONG_PTR,
>  			      MCFGR_AWCACHE_CACH | MCFGR_AWCACHE_BUFF |
>  			      MCFGR_WDENABLE | MCFGR_LARGE_BURST |
> -			      (sizeof(dma_addr_t) == sizeof(u64) ?
> +			      (caam_ptr_sz == sizeof(u64) ?
>  			       MCFGR_LONG_PTR : 0));
>  
Before this in caam_probe() there's:
	if (comp_params & CTPR_MS_PS && rd_reg32(&ctrl->mcr) & MCFGR_LONG_PTR)
		caam_ptr_sz = sizeof(u64);
	else
		caam_ptr_sz = sizeof(u32);

thus caam_ptr_sz is determined based on MCFGR:
MCFGR[PS] ==> caam_ptr_sz
so it no longer makes sense to reconfigure MCFGR based on caam_ptr_sz:
caam_ptr_sz ==> MCFGR[PS]

The short-term fix would be to no longer touch MCFGR[PS], which is in line with
commit a1cf573ee95 ("crypto: caam - select DMA address size at runtime")

Note however there's a small issue with the logic of kernel relying on
MCFGR[PS] being previously configured (by U-boot etc.), see
commit 39eaf759466f ("crypto: caam - fix pointer size for AArch64 boot loader, AArch32 kernel")

In the long run, IMO the proper thing for the driver to do is to rely,
whenever possible, on DT ("dma-ranges" property) / dma_mask provided by
the device driver infrastructure, and not on MCFGR[PS]:
[ DT "dma-ranges" ==> ] dma_mask ==> MCFGR[PS], caam_ptr_sz 
but that would probably involve too many changes for getting into -stable.

Horia