| #
3bf3710e |
|
02-Jun-2021 |
Thomas Hellström <thomas.hellstrom@linux.intel.com> |
drm/ttm: Add a generic TTM memcpy move for page-based iomem
The internal ttm_bo_util memcpy uses ioremap functionality, and while it
probably might be possible to use it for copying in- and out of
sglist represented io memory, using io_mem_reserve() / io_mem_free()
callbacks, that would cause problems with fault().
Instead, implement a method mapping page-by-page using kmap_local()
semantics. As an additional benefit we then avoid the occasional global
TLB flushes of ioremap() and consuming ioremap space, elimination of a
critical point of failure and with a slight change of semantics we could
also push the memcpy out async for testing and async driver development
purposes.
A special linear iomem iterator is introduced internally to mimic the
old ioremap behaviour for code-paths that can't immediately be ported
over. This adds to the code size and should be considered a temporary
solution.
Looking at the code we have a lot of checks for iomap tagged pointers.
Ideally we should extend the core memremap functions to also accept
uncached memory and kmap_local functionality. Then we could strip a
lot of code.
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210602083818.241793-4-thomas.hellstrom@linux.intel.com
|
| #
d099fc8f |
|
22-Oct-2020 |
Christian König <christian.koenig@amd.com> |
drm/ttm: new TT backend allocation pool v3
This replaces the spaghetti code in the two existing page pools.
First of all depending on the allocation size it is between 3 (1GiB) and
5 (1MiB) times faster than the old implementation.
It makes better use of buddy pages to allow for larger physical contiguous
allocations which should result in better TLB utilization at least for
amdgpu.
Instead of a completely braindead approach of filling the pool with one
CPU while another one is trying to shrink it we only give back freed
pages.
This also results in much less locking contention and a trylock free MM
shrinker callback, so we can guarantee that pages are given back to the
system when needed.
Downside of this is that it takes longer for many small allocations until
the pool is filled up. We could address this, but I couldn't find an use
case where this actually matters. We also don't bother freeing large
chunks of pages any more since the CPU overhead in that path isn't really
that important.
The sysfs files are replaced with a single module parameter, allowing
users to override how many pages should be globally pooled in TTM. This
unfortunately breaks the UAPI slightly, but as far as we know nobody ever
depended on this.
Zeroing memory coming from the pool was handled inconsistently. The
alloc_pages() based pool was zeroing it, the dma_alloc_attr() based one
wasn't. For now the new implementation isn't zeroing pages from the pool
either and only sets the __GFP_ZERO flag when necessary.
The implementation has only 768 lines of code compared to the over 2600
of the old one, and also allows for saving quite a bunch of code in the
drivers since we don't need specialized handling there any more based on
kernel config.
Additional to all of that there was a neat bug with IOMMU, coherent DMA
mappings and huge pages which is now fixed in the new code as well.
v2: make ttm_pool_apply_caching static as reported by the kernel bot, add
some more checks
v3: fix some more checkpatch.pl warnings
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Madhav Chauhan <madhav.chauhan@amd.com>
Tested-by: Huang Rui <ray.huang@amd.com>
Link: https://patchwork.freedesktop.org/patch/397080/?series=83051&rev=1
|