nvidia_x11: install optimus ICDs#17
Merged
eadwu merged 1 commit intonvidia/prime-render-offloadfrom Sep 8, 2019
unknown repository
Merged
nvidia_x11: install optimus ICDs#17eadwu merged 1 commit intonvidia/prime-render-offloadfrom unknown repository
eadwu merged 1 commit intonvidia/prime-render-offloadfrom
unknown repository
Conversation
ghost
mentioned this pull request
10 tasks
Author
|
cc @eadwu |
Author
|
@eadwu Thank you! |
ghost
deleted the
eadwu
pushed a commit
that referenced
this pull request
Apr 3, 2025
fluent-bit 3.2.7, 3.2.8 and 3.2.9 are segfaulting when used in combination with the systemd input. Lets revert to 3.2.6 for now. Upstream bug: fluent/fluent-bit#10139 Note that fluent-bit-3.2.7 fixes two high CVEs which we are now reintroducing. However they are only exploitable if you are using the OpenTelemetry input or the Prometheus Remote Write input. OpenTelemetry input: [CVE-2024-50609](https://nvd.nist.gov/vuln/detail/CVE-2024-50609) Prometheus Remote Write input: [CVE-2024-50608](https://nvd.nist.gov/vuln/detail/CVE-2024-50608) The problem is as follows: 3.2.7 started vendoring a copy of `libzstd` in tree and statically linking against it. Also, the fluent-bit binary exports the symbols of static libraries it links against. This is a problem because `libzstd` gets `dlopen()`ed by `libsystemd` when enumerating the journal (as journal logs are zstd compressed). and `libzstd` in Nixpkgs is built with `-DZSTD_LEGACY_SUPPORT=0` which causes `struct ZSTD_DCtx` to be 16 bytes smaller than without this flag https://github.com/facebook/zstd/blob/dev/lib/decompress/zstd_decompress_internal.h#L183-L187 `libsystemd` calls [`sym_ZSTD_createDCtx()`](https://github.com/systemd/systemd/blob/1e79a2923364b65fc9f347884dd5b9b2087f6e32/src/basic/compress.c#L480) which calls the function pointer returned by `dlsym()` which is calling into the `libzstd` that comes with `nixpkgs` and thus allocates a struct that is 16 bytes smaller. Later then `sym_ZSTD_freeDCtx()` is called. However because fluent-bit has `zstd` in its global symbol table, any functions that `sym_ZSTD_freeDCtx()` calls will be calls to the functions in the vendored fluent-bit version of the library which expects the larger struct. This then causes enough heap corruption to cause a segfault. E.g. the subsequent calls to `ZSTD_clearDict(dctx)` and `ZSTD_customFree(dctx->inBuff)` in https://github.com/facebook/zstd/blob/dev/lib/decompress/zstd_decompress.c#L324 will be working on a struct that is 16 bytes smaller than the one that was allocated by `libsystemd` and will cause a segfault at some point and thus are probably modifying pieces of memory that they shouldn't (gdb) bt #0 0x00007f10e7e9916c in __pthread_kill_implementation () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #1 0x00007f10e7e40e86 in raise () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #2 0x00007f10e7e2893a in abort () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #3 0x000000000046a938 in flb_signal_handler () #4 <signal handler called> #5 0x00007f10e7ea42b7 in unlink_chunk.isra () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #6 0x00007f10e7ea45cd in _int_free_create_chunk () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #7 0x00007f10e7ea5a1c in _int_free_merge_chunk () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #8 0x00007f10e7ea5dc9 in _int_free () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #9 0x00007f10e7ea8613 in free () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 #10 0x00007f10e80ad3b5 in ZSTD_freeDCtx () from /nix/store/wy0slah6yvchgra8nhp6vgrqa6ay72cq-zstd-1.5.6/lib/libzstd.so.1 #11 0x00007f10e8c90f6b in decompress_blob_zstd () from /nix/store/b2cfj7yk3wfg1jdwjzim7306hvsc5gnl-systemd-257.3/lib/libsystemd.so.0 #12 0x00007f10e8bf0efe in journal_file_data_payload () from /nix/store/b2cfj7yk3wfg1jdwjzim7306hvsc5gnl-systemd-257.3/lib/libsystemd.so.0 #13 0x00007f10e8c00f74 in sd_journal_enumerate_data () from /nix/store/b2cfj7yk3wfg1jdwjzim7306hvsc5gnl-systemd-257.3/lib/libsystemd.so.0 #14 0x00000000004eae2f in in_systemd_collect () #15 0x00000000004eb5a0 in in_systemd_collect_archive () #16 0x000000000047aa18 in flb_input_collector_fd () #17 0x0000000000495223 in flb_engine_start () NixOS#18 0x000000000046f304 in flb_lib_worker () NixOS#19 0x00007f10e7e972e3 in start_thread () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 NixOS#20 0x00007f10e7f1b2fc in __clone3 () from /nix/store/rmy663w9p7xb202rcln4jjzmvivznmz8-glibc-2.40-66/lib/libc.so.6 Reverts 7310ab3 Reverts 4fbc6cf
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Now Vulkan part works. No idea what is wrong with OpenGL
Both work fine. Background is different so its easy to distinct these two
That worked too. I'm not sure why nvidia is darker...
Loaded
Idle