mtmd : fix memory leak in mtmd_helper_eval_chunk_single #13961
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
DangNhutNguyen
approved these changes
Jun 2, 2025
evverx
reviewed
Jun 2, 2025
| @@ -311,6 +311,7 @@ int32_t mtmd_helper_eval_chunk_single(mtmd_context * ctx, | |||
| GGML_ABORT("chunk type not supported"); | |||
| } | |||
|
|
|||
| llama_batch_free(text_batch); | |||
There was a problem hiding this comment.
FWIW It seems to fix the following direct leaks
Direct leak of 49176 byte(s) in 3 object(s) allocated from:
#0 0x0000004abdc8 in malloc (/llama.cpp/build/bin/llama-mtmd-cli+0x4abdc8) (BuildId: d0529438afdc212f2f3e006e74cba8cc230cfcc6)
#1 0x7fcb42e3a64b in llama_batch_init /llama.cpp/src/llama-batch.cpp:358:40
#2 0x7fcb4337b518 in mtmd_helper_eval_chunk_single /llama.cpp/tools/mtmd/mtmd-helper.cpp:255:30
#3 0x7fcb4337bcab in mtmd_helper_eval_chunks /llama.cpp/tools/mtmd/mtmd-helper.cpp:335:23
#4 0x0000004f3aa6 in eval_message(mtmd_cli_context&, common_chat_msg&, bool) /llama.cpp/tools/mtmd/mtmd-cli.cpp:222:9
#5 0x0000004f22fc in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:300:13
#6 0x7fcb425a95f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#7 0x7fcb425a96a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#8 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: d0529438afdc212f2f3e006e74cba8cc230cfcc6)
...but with this PR applied ASan still reports memory leaks in llama-mtmd-cli:
SUMMARY: AddressSanitizer: 654872 byte(s) leaked in 2096 allocation(s).
Direct leak of 16392 byte(s) in 1 object(s) allocated from:
#0 0x0000004abdc8 in malloc (/llama.cpp/build/bin/llama-mtmd-cli+0x4abdc8) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
#1 0x7f748b1d464b in llama_batch_init /llama.cpp/src/llama-batch.cpp:358:40
#2 0x0000004f8d9e in mtmd_cli_context::mtmd_cli_context(common_params&) /llama.cpp/tools/mtmd/mtmd-cli.cpp:93:17
#3 0x0000004f1ea1 in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:259:22
#4 0x7f748a9435f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#5 0x7f748a9436a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#6 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
Direct leak of 8192 byte(s) in 1 object(s) allocated from:
#0 0x0000004abdc8 in malloc (/llama.cpp/build/bin/llama-mtmd-cli+0x4abdc8) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
#1 0x7f748b1d461e in llama_batch_init /llama.cpp/src/llama-batch.cpp:357:40
#2 0x0000004f8d9e in mtmd_cli_context::mtmd_cli_context(common_params&) /llama.cpp/tools/mtmd/mtmd-cli.cpp:93:17
#3 0x0000004f1ea1 in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:259:22
#4 0x7f748a9435f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#5 0x7f748a9436a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#6 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
Direct leak of 8192 byte(s) in 1 object(s) allocated from:
#0 0x0000004abdc8 in malloc (/llama.cpp/build/bin/llama-mtmd-cli+0x4abdc8) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
#1 0x7f748b1d45fa in llama_batch_init /llama.cpp/src/llama-batch.cpp:356:40
#2 0x0000004f8d9e in mtmd_cli_context::mtmd_cli_context(common_params&) /llama.cpp/tools/mtmd/mtmd-cli.cpp:93:17
#3 0x0000004f1ea1 in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:259:22
#4 0x7f748a9435f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#5 0x7f748a9436a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#6 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
Direct leak of 8192 byte(s) in 1 object(s) allocated from:
#0 0x0000004abdc8 in malloc (/llama.cpp/build/bin/llama-mtmd-cli+0x4abdc8) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
#1 0x7f748b1d45cb in llama_batch_init /llama.cpp/src/llama-batch.cpp
#2 0x0000004f8d9e in mtmd_cli_context::mtmd_cli_context(common_params&) /llama.cpp/tools/mtmd/mtmd-cli.cpp:93:17
#3 0x0000004f1ea1 in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:259:22
#4 0x7f748a9435f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#5 0x7f748a9436a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#6 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
Direct leak of 2048 byte(s) in 1 object(s) allocated from:
#0 0x0000004abdc8 in malloc (/llama.cpp/build/bin/llama-mtmd-cli+0x4abdc8) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
#1 0x7f748b1d46e0 in llama_batch_init /llama.cpp/src/llama-batch.cpp:364:40
#2 0x0000004f8d9e in mtmd_cli_context::mtmd_cli_context(common_params&) /llama.cpp/tools/mtmd/mtmd-cli.cpp:93:17
#3 0x0000004f1ea1 in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:259:22
#4 0x7f748a9435f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#5 0x7f748a9436a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#6 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
Direct leak of 416 byte(s) in 1 object(s) allocated from:
#0 0x0000004f07c1 in operator new(unsigned long) (/llama.cpp/build/bin/llama-mtmd-cli+0x4f07c1) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)
#1 0x0000007f6a4f in common_sampler_init(llama_model const*, common_params_sampling const&) /llama.cpp/common/sampling.cpp:216:21
#2 0x0000004f1fba in main /llama.cpp/tools/mtmd/mtmd-cli.cpp:264:36
#3 0x7f748a9435f4 in __libc_start_call_main (/lib64/libc.so.6+0x35f4) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#4 0x7f748a9436a7 in __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x36a7) (BuildId: 2b3c02fe7e4d3811767175b6f323692a10a4e116)
#5 0x0000004082b4 in _start (/llama.cpp/build/bin/llama-mtmd-cli+0x4082b4) (BuildId: 693cd788cb564003831ab992f5b80d129be780e8)There was a problem hiding this comment.
It should be fixed in 6aaefb7
Could you give it a try?
There was a problem hiding this comment.
Looks like ASan no longer complains. Thanks!
|
FWIW, adding this line does fix the leak in my project without any noticeable side effects. |
ggerganov
approved these changes
Jun 2, 2025
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix #13958
Maybe we should have a C++ wrapper in
llama-cpp.hforllama_batch?