linux-kernel - Re: [RFC PATCH v2 06/11] riscv: pgtable: Add DMA

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:   Sun, 06 Jun 2021 20:22:14 +0300
From:   Nick Kossifidis <mick@....forth.gr>
To:     guoren@...nel.org
Cc:     anup.patel@....com, palmerdabbelt@...gle.com, arnd@...db.de,
        wens@...e.org, maxime@...no.tech, drew@...gleboard.org,
        liush@...winnertech.com, lazyparser@...il.com, wefu@...hat.com,
        linux-riscv@...ts.infradead.org, linux-kernel@...r.kernel.org,
        linux-arch@...r.kernel.org, linux-sunxi@...ts.linux.dev,
        Guo Ren <guoren@...ux.alibaba.com>,
        Christoph Hellwig <hch@....de>
Subject: Re: [RFC PATCH v2 06/11] riscv: pgtable: Add DMA_COHERENT with custom
 PTE attributes

Στις 2021-06-06 12:04, guoren@...nel.org έγραψε:
> From: Guo Ren <guoren@...ux.alibaba.com>
> 
> The dma-noncoherent SOCs need different virtual memory mappings
> with different attributes:
>  - noncached + Strong Order (for IO/DMA descriptor)
>  - noncached + Weak Order (for writecombine usage, eg: frame
>    buffer)
> 
> All above base on PTE attributes by MMU hardware. That means
> address attributes are determined by PTE entry, not PMA. RISC-V
> soc vendors have defined their own custom PTE attributes for
> dma-noncoherency.
> 

This patch violates the Privilege Spec section 4.4.1 that clearly 
states:

"Bits63–54 are reserved for future standard use and must be zeroed by 
software for forward compatibility"

Standard use means that valid values can only be defined by the Priv. 
Spec, not by the vendor (otherwise they'd be marked as "custom use" or 
"platform use"), and since they "must" be zeroed by software we 'll be 
violating the Privilege Spec if we do otherwise.