linux-kernel - RE: [PATCH v6 0/6] Fix double allocation in swiotlb

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: Fri, 8 Mar 2024 19:21:23 +0000
From: Michael Kelley <mhklinux@...look.com>
To: Will Deacon <will@...nel.org>, "linux-kernel@...r.kernel.org"
	<linux-kernel@...r.kernel.org>
CC: "kernel-team@...roid.com" <kernel-team@...roid.com>,
	"iommu@...ts.linux.dev" <iommu@...ts.linux.dev>, Christoph Hellwig
	<hch@....de>, Marek Szyprowski <m.szyprowski@...sung.com>, Robin Murphy
	<robin.murphy@....com>, Petr Tesarik <petr.tesarik1@...wei-partners.com>,
	Dexuan Cui <decui@...rosoft.com>, Nicolin Chen <nicolinc@...dia.com>
Subject: RE: [PATCH v6 0/6] Fix double allocation in swiotlb_alloc()

From: Will Deacon <will@...nel.org> Sent: Friday, March 8, 2024 7:28 AM
> 

[snip]

> 
> Changes since v5 include:
> 
>   - Rework the final patch to preserve page-alignment for streaming
>     requests without a DMA alignment mask
> 
>   - Added Reviewed-by tags from Michael
> 
> Cheers,
> 
> Will
> 

I've tested the full v6 of this series on an ARM64 VM in the Azure
public cloud with 64K page size, and swiotlb=force on the kernel
boot line.  The path exercised is the DMA Direct path on a synthetic
SCSI disk device with min_align_mask = 0xFFF.  I have a standalone
user-space test program using writev() to generate direct disk
I/Os with a variety of memory buffer alignments and physically
contiguous extents.  It's designed to test the paths through the
iovec level, scatter/gather lists, and the driver DMA descriptors
to make sure all is working correctly.

Prior to this patch series, some tests were failing with 64K page
size when swiotlb=force, but they are all working now.

Tested-by: Michael Kelley <mhklinux@...look.com>