linux-kernel - dma-buf CPU access

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

Message-ID: <e2571723-1b92-db19-8ca8-06f358363453@intel.com>
Date:   Fri, 9 Feb 2018 23:25:44 +0200
From:   Alexey Skidanov <alexey.skidanov@...el.com>
To:     sumit.semwal@...aro.org, linux-media@...r.kernel.org,
        linux-kernel@...r.kernel.org
Cc:     Laura Abbott <labbott@...hat.com>
Subject: dma-buf CPU access

Hi,

According to the CPU access instructions in the dma-buf.c, I understand
that on 32 bit platforms, due to the limited number of persistent kernel
mappings or limited vmalloc range, the CPU access pattern should look
like this:

dma_buf_start_cpu_access();

void *ptr = dma_buf_kmap();

/*Access the page pointed by ptr*/

dma_buf_kunmap(ptr);

dma_buf_end_cpu_access();


or


dma_buf_start_cpu_access();

void *ptr = dma_buf_vmap();

/*Access the buffer pointed by ptr*/

dma_buf_vunmap(ptr);

dma_buf_end_cpu_access();


But on 64 bit platforms, there are no such restrictions (there is no
highmem at all). So, frequently accessed buffer, may be mapped once, but
every access should be surrounded by :

dma_buf_start_cpu_access();

/*Access the buffer pointed by ptr*/

dma_buf_end_cpu_access()

Am I correct?

Thanks,
Alexey