]> git.baikalelectronics.ru Git - kernel.git/commit
Revert "iommu/dma: Add config for PCI SAC address trick"
authorLinus Torvalds <torvalds@linux-foundation.org>
Sat, 6 Aug 2022 20:24:56 +0000 (13:24 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 6 Aug 2022 20:24:56 +0000 (13:24 -0700)
commitce772fcf94e5c9383c96d0720c3c9a86ac4637cf
tree116fe5e763f3531379a4730ad6e9efbb8aa69433
parent3886cbaff1138e9a956f6945a248933da4d2684c
Revert "iommu/dma: Add config for PCI SAC address trick"

This reverts commit 81903e89f8e84459ec9b9264447c8239df423d17.

It turns out that it was hopelessly naive to think that this would work,
considering that we've always done this.  The first machine I actually
tested this on broke at bootup, getting to

    Reached target cryptsetup.target - Local Encrypted Volumes.

and then hanging.  It's unclear what actually fails, since there's a lot
else going on around that time (eg amdgpu probing also happens around
that same time, but it could be some other random init thing that didn't
complete earlier and just caused the boot to hang at that point).

The expectations that we should default to some unsafe and untested mode
seems entirely unfounded, and the belief that this wouldn't affect
modern systems is clearly entirely false.  The machine in question is
about two years old, so it's not exactly shiny, but it's also not some
dusty old museum piece PDP-11 in a closet.

Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: John Garry <john.garry@huawei.com>
Cc: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/iommu/Kconfig
drivers/iommu/dma-iommu.c