tjh.dev / kernel
Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

Revert "iommu/dma: Add config for PCI SAC address trick"

This reverts commit 4bf7fda4dce22214c70c49960b1b6438e6260b67.

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>

Linus Torvalds af3e9579 20cf903a

+1 -27
-26
drivers/iommu/Kconfig
··· 144 144 select IRQ_MSI_IOMMU 145 145 select NEED_SG_DMA_LENGTH 146 146 147 - config IOMMU_DMA_PCI_SAC 148 - bool "Enable 64-bit legacy PCI optimisation by default" 149 - depends on IOMMU_DMA 150 - help 151 - Enable by default an IOMMU optimisation for 64-bit legacy PCI devices, 152 - wherein the DMA API layer will always first try to allocate a 32-bit 153 - DMA address suitable for a single address cycle, before falling back 154 - to allocating from the device's full usable address range. If your 155 - system has 64-bit legacy PCI devices in 32-bit slots where using dual 156 - address cycles reduces DMA throughput significantly, this may be 157 - beneficial to overall performance. 158 - 159 - If you have a modern PCI Express based system, this feature mostly just 160 - represents extra overhead in the allocation path for no practical 161 - benefit, and it should usually be preferable to say "n" here. 162 - 163 - However, beware that this feature has also historically papered over 164 - bugs where the IOMMU address width and/or device DMA mask is not set 165 - correctly. If device DMA problems and IOMMU faults start occurring 166 - after disabling this option, it is almost certainly indicative of a 167 - latent driver or firmware/BIOS bug, which would previously have only 168 - manifested with several gigabytes worth of concurrent DMA mappings. 169 - 170 - If this option is not set, the feature can still be re-enabled at 171 - boot time with the "iommu.forcedac=0" command-line argument. 172 - 173 147 # Shared Virtual Addressing 174 148 config IOMMU_SVA 175 149 bool
+1 -1
drivers/iommu/dma-iommu.c
··· 69 69 }; 70 70 71 71 static DEFINE_STATIC_KEY_FALSE(iommu_deferred_attach_enabled); 72 - bool iommu_dma_forcedac __read_mostly = !IS_ENABLED(CONFIG_IOMMU_DMA_PCI_SAC); 72 + bool iommu_dma_forcedac __read_mostly; 73 73 74 74 static int __init iommu_dma_forcedac_setup(char *str) 75 75 {