| Age | Commit message (Collapse) | Author |
|---|
|
Patch series "Memory allocation profiling", v6.
Overview:
Low overhead [1] per-callsite memory allocation profiling. Not just for
debug kernels, overhead low enough to be deployed in production.
Example output:
root@moria-kvm:~# sort -rn /proc/allocinfo
127664128 31168 mm/page_ext.c:270 func:alloc_page_ext
56373248 4737 mm/slub.c:2259 func:alloc_slab_page
14880768 3633 mm/readahead.c:247 func:page_cache_ra_unbounded
14417920 3520 mm/mm_init.c:2530 func:alloc_large_system_hash
13377536 234 block/blk-mq.c:3421 func:blk_mq_alloc_rqs
11718656 2861 mm/filemap.c:1919 func:__filemap_get_folio
9192960 2800 kernel/fork.c:307 func:alloc_thread_stack_node
4206592 4 net/netfilter/nf_conntrack_core.c:2567 func:nf_ct_alloc_hashtable
4136960 1010 drivers/staging/ctagmod/ctagmod.c:20 [ctagmod] func:ctagmod_start
3940352 962 mm/memory.c:4214 func:alloc_anon_folio
2894464 22613 fs/kernfs/dir.c:615 func:__kernfs_new_node
...
Usage:
kconfig options:
- CONFIG_MEM_ALLOC_PROFILING
- CONFIG_MEM_ALLOC_PROFILING_ENABLED_BY_DEFAULT
- CONFIG_MEM_ALLOC_PROFILING_DEBUG
adds warnings for allocations that weren't accounted because of a
missing annotation
sysctl:
/proc/sys/vm/mem_profiling
Runtime info:
/proc/allocinfo
Notes:
[1]: Overhead
To measure the overhead we are comparing the following configurations:
(1) Baseline with CONFIG_MEMCG_KMEM=n
(2) Disabled by default (CONFIG_MEM_ALLOC_PROFILING=y &&
CONFIG_MEM_ALLOC_PROFILING_BY_DEFAULT=n)
(3) Enabled by default (CONFIG_MEM_ALLOC_PROFILING=y &&
CONFIG_MEM_ALLOC_PROFILING_BY_DEFAULT=y)
(4) Enabled at runtime (CONFIG_MEM_ALLOC_PROFILING=y &&
CONFIG_MEM_ALLOC_PROFILING_BY_DEFAULT=n && /proc/sys/vm/mem_profiling=1)
(5) Baseline with CONFIG_MEMCG_KMEM=y && allocating with __GFP_ACCOUNT
(6) Disabled by default (CONFIG_MEM_ALLOC_PROFILING=y &&
CONFIG_MEM_ALLOC_PROFILING_BY_DEFAULT=n) && CONFIG_MEMCG_KMEM=y
(7) Enabled by default (CONFIG_MEM_ALLOC_PROFILING=y &&
CONFIG_MEM_ALLOC_PROFILING_BY_DEFAULT=y) && CONFIG_MEMCG_KMEM=y
Performance overhead:
To evaluate performance we implemented an in-kernel test executing
multiple get_free_page/free_page and kmalloc/kfree calls with allocation
sizes growing from 8 to 240 bytes with CPU frequency set to max and CPU
affinity set to a specific CPU to minimize the noise. Below are results
from running the test on Ubuntu 22.04.2 LTS with 6.8.0-rc1 kernel on
56 core Intel Xeon:
kmalloc pgalloc
(1 baseline) 6.764s 16.902s
(2 default disabled) 6.793s (+0.43%) 17.007s (+0.62%)
(3 default enabled) 7.197s (+6.40%) 23.666s (+40.02%)
(4 runtime enabled) 7.405s (+9.48%) 23.901s (+41.41%)
(5 memcg) 13.388s (+97.94%) 48.460s (+186.71%)
(6 def disabled+memcg) 13.332s (+97.10%) 48.105s (+184.61%)
(7 def enabled+memcg) 13.446s (+98.78%) 54.963s (+225.18%)
Memory overhead:
Kernel size:
text data bss dec diff
(1) 26515311 18890222 17018880 62424413
(2) 26524728 19423818 16740352 62688898 264485
(3) 26524724 19423818 16740352 62688894 264481
(4) 26524728 19423818 16740352 62688898 264485
(5) 26541782 18964374 16957440 62463596 39183
Memory consumption on a 56 core Intel CPU with 125GB of memory:
Code tags: 192 kB
PageExts: 262144 kB (256MB)
SlabExts: 9876 kB (9.6MB)
PcpuExts: 512 kB (0.5MB)
Total overhead is 0.2% of total memory.
Benchmarks:
Hackbench tests run 100 times:
hackbench -s 512 -l 200 -g 15 -f 25 -P
baseline disabled profiling enabled profiling
avg 0.3543 0.3559 (+0.0016) 0.3566 (+0.0023)
stdev 0.0137 0.0188 0.0077
hackbench -l 10000
baseline disabled profiling enabled profiling
avg 6.4218 6.4306 (+0.0088) 6.5077 (+0.0859)
stdev 0.0933 0.0286 0.0489
stress-ng tests:
stress-ng --class memory --seq 4 -t 60
stress-ng --class cpu --seq 4 -t 60
Results posted at: https://evilpiepirate.org/~kent/memalloc_prof_v4_stress-ng/
[2] https://lore.kernel.org/all/20240306182440.2003814-1-surenb@google.com/
This patch (of 37):
The next patch drops vmalloc.h from a system header in order to fix a
circular dependency; this adds it to all the files that were pulling it in
implicitly.
[kent.overstreet@linux.dev: fix arch/alpha/lib/memcpy.c]
Link: https://lkml.kernel.org/r/20240327002152.3339937-1-kent.overstreet@linux.dev
[surenb@google.com: fix arch/x86/mm/numa_32.c]
Link: https://lkml.kernel.org/r/20240402180933.1663992-1-surenb@google.com
[kent.overstreet@linux.dev: a few places were depending on sizes.h]
Link: https://lkml.kernel.org/r/20240404034744.1664840-1-kent.overstreet@linux.dev
[arnd@arndb.de: fix mm/kasan/hw_tags.c]
Link: https://lkml.kernel.org/r/20240404124435.3121534-1-arnd@kernel.org
[surenb@google.com: fix arc build]
Link: https://lkml.kernel.org/r/20240405225115.431056-1-surenb@google.com
Link: https://lkml.kernel.org/r/20240321163705.3067592-1-surenb@google.com
Link: https://lkml.kernel.org/r/20240321163705.3067592-2-surenb@google.com
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Tested-by: Kees Cook <keescook@chromium.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Alex Gaynor <alex.gaynor@gmail.com>
Cc: Alice Ryhl <aliceryhl@google.com>
Cc: Andreas Hindborg <a.hindborg@samsung.com>
Cc: Benno Lossin <benno.lossin@proton.me>
Cc: "Björn Roy Baron" <bjorn3_gh@protonmail.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Gary Guo <gary@garyguo.net>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Wedson Almeida Filho <wedsonaf@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
This adds support for the
SVGA_3D_CMD_DX_SET_VS/PS/GS/HS/DS/CS_CONSTANT_BUFFER_OFFSET commands (which only update
the offset, but don't rebind the buffer), which saves some overhead.
Signed-off-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Signed-off-by: Zack Rusin <zackr@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-11-zack@kde.org
|
|
If the host supports SVGA3D_DEVCAP_GL43, we can handle 64 instead of
just 8 UAVs.
Based on a patch from Roland Scheidegger <sroland@vmware.com>.
Signed-off-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-9-zack@kde.org
|
|
vmwgfx shared very elaborate memory accounting with ttm. It was moved
from ttm to vmwgfx in change
f07069da6b4c ("drm/ttm: move memory accounting into vmwgfx v4")
but because of complexity it was hard to maintain. Some parts of the code
weren't freeing memory correctly and some were missing accounting all
together. While those would be fairly easy to fix the fundamental reason
for memory accounting in the driver was the ability to invoke shrinker
which is part of TTM code as well (with support for unified memory
hopefully coming soon).
That meant that vmwgfx had a lot of code that was either unused or
duplicating code from TTM. Removing this code also prevents excessive
calls to global swapout which were common during memory pressure
because both vmwgfx and TTM would invoke the shrinker when memory
usage reached half of RAM.
Fixes: f07069da6b4c ("drm/ttm: move memory accounting into vmwgfx v4")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-2-zack@kde.org
|
|
Historically our device headers have been forked versions of the
internal device headers, this has made maintaining them a bit
of a burden. To fix the situation, going forward, the device headers
will be verbatim copies of the internal headers.
To do that the driver code has to be adapted to use pristine
device headers. This will make future update to the device
headers trivial and automatic.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210615182336.995192-2-zackr@vmware.com
|
|
Fix some minor issues that Coverity spotted in the code. None
of that are serious but they're all valid concerns so fixing
them makes sense.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210609172307.131929-5-zackr@vmware.com
|
|
SVGA3 is the next version of our PCI device. Some of the changes
include using MMIO for register accesses instead of ioports,
deprecating the FIFO MMIO and removing a lot of the old and
legacy functionality. SVGA3 doesn't support guest backed
objects right now so everything except 3D is working.
v2: Fixes all the static analyzer warnings
Signed-off-by: Zack Rusin <zackr@vmware.com>
Cc: Martin Krastev <krastevm@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210505191007.305872-1-zackr@vmware.com
|
|
remove others
Fixes the following W=1 kernel build warning(s):
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:340: warning: Function parameter or member 'shader_slot' not described in 'vmw_binding_add'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:340: warning: Function parameter or member 'slot' not described in 'vmw_binding_add'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:376: warning: Function parameter or member 'from' not described in 'vmw_binding_transfer'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:498: warning: Function parameter or member 'to' not described in 'vmw_binding_state_commit'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:498: warning: Excess function parameter 'ctx' description in 'vmw_binding_state_commit'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:498: warning: Excess function parameter 'scrubbed' description in 'vmw_binding_state_commit'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:520: warning: Function parameter or member 'cbs' not described in 'vmw_binding_rebind_all'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:520: warning: Excess function parameter 'ctx' description in 'vmw_binding_rebind_all'
drivers/gpu/drm/vmwgfx/vmwgfx_binding.c:795: warning: Function parameter or member 'shader_slot' not described in 'vmw_emit_set_sr'
Cc: VMware Graphics <linux-graphics-maintainer@vmware.com>
Cc: Roland Scheidegger <sroland@vmware.com>
Cc: Zack Rusin <zackr@vmware.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Zack Rusin <zackr@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210115181601.3432599-5-lee.jones@linaro.org
|
|
Lets try to cleanup the usage of the term FIFO which we used for
both our MMIO based cmd queue processing and for general
command processing which could have been using command buffers
interface. We're going to rename the functions which are processing
commands (and work either via MMIO or command buffers) as _cmd_
and functions which operate on the MMIO based commands as FIFO
to match the SVGA device naming.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/414044/?series=85516&rev=2
|
|
With SM5 capability a new version of streamoutput is supported by device
which need backing mob and a new field. With this change the new command
is supported in command buffer.
v2: Also track streamoutput context binding in binding manager.
v3: Track only one streamoutput as only one can be set to context.
v4: Fix comment typos
Signed-off-by: Deepak Rawat <drawat.floss@gmail.com>
Signed-off-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Thomas Hellström (VMware) <thomas_os@shipmail.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Roland Scheidegger <sroland@vmware.com>
|
|
Previous name vmw_ctx_bindinfo_so is misleading because it actually
represent so target and stream output is a new resource type that needs
tracking for SM5 capable device. Also rename binding type enum and
internal functions to reflect these belongs to so targets.
Signed-off-by: Deepak Rawat <drawat.floss@gmail.com>
Reviewed-by: Thomas Hellström (VMware) <thomas_os@shipmail.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Roland Scheidegger <sroland@vmware.com>
|
|
Virtual device now support new commands to manage unordered access
views. Allow them as part of user-space command buffer. This involves
adding UA view cotable, binding tracker info, new view type and command
verifier functions.
v2: fix comment typo
v3: style fixes (don't use deprecated PTR_RET)
Signed-off-by: Deepak Rawat <drawat.floss@gmail.com>
Signed-off-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Thomas Hellström (VMware) <thomas_os@shipmail.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Roland Scheidegger <sroland@vmware.com>
|
|
Virtual device now supports new shader types, allow them as valid shader
type in command buffer. Also add per shader bind info in binding manager
state for new shader type.
Signed-off-by: Deepak Rawat <drawat.floss@gmail.com>
Reviewed-by: Thomas Hellström (VMware) <thomas_os@shipmail.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Roland Scheidegger <sroland@vmware.com>
|
|
Whenever FIFO allocation fails an error message is printed to dmesg.
Since this is common operation a lot of similar messages are scattered
everywhere. Use preprocessor macro to remove this cluttering.
Signed-off-by: Deepak Rawat <drawat@vmware.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
|
|
Currently we flag resources as dirty (GPU contents not yet read back to
the backing MOB) whenever they have been part of a command stream.
Obviously many resources can't be dirty and others can only be dirty when
written to by the GPU. That is when they are either bound to the context as
render-targets, depth-stencil, copy / clear destinations and
stream-output targets, or similarly when there are corresponding views into
them.
So mark resources dirty only in these special cases. Context- and cotable
resources are always marked dirty when referenced.
This is important for upcoming emulated coherent memory, since we can avoid
issuing automatic readbacks to non-dirty resources when the CPU tries to
access part of the backing MOB.
Testing: Unigine Heaven with max GPU memory set to 256MB resulting in
heavy resource thrashing.
---
v2: Addressed review comments by Deepak Rawat.
v3: Added some documentation
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
|
|
This is dual licensed under GPL-2.0 or MIT.
vmwgfx_msg.h is the odd one out that is GPL-2.0+ or MIT.
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Dirk Hohndel (VMware) <dirk@hohndel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180506231626.115996-9-dirk@hohndel.org
|
|
forward the operation context to ttm_mem_global_alloc as well, and the
ultimate goal is swapout enablement for reserved BOs
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Roger He <Hongbo.He@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Initial DX support.
Co-authored with Sinclair Yeh, Charmaine Lee and Jakob Bornecrantz.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Sinclair Yeh <syeh@vmware.com>
Signed-off-by: Charmaine Lee <charmainel@vmware.com>
|
