Possible leaks when using foreign export.
When a Haskell function exported with foreign export is called from outside Haskell a new Task may be allocated for it. There will be one Task allocated for each OS thread calling into Haskell (this happens in rts_lock()). This Task is, by design, never deallocated. This works fine with a small number of threads, but if the external caller generates a lot of threads that die the Haskell RTS will just get more and more Task objects that are never freed. The Task also contains a condition variable and a mutex, which incurs a HANDLE leak on Windows.
I don't see an elegant fix to this, but the attached code remembers the OS thread for "incall" Task, and every so often (every 100 Tasks created) if will scan the Tasks looking for ones where the OS thread is dead, and remove these. The fix only works on Windows, I'm afraid.
Trac metadata
| Trac field | Value |
|---|---|
| Version | 7.6.3 |
| Type | Bug |
| TypeOfFailure | OtherFailure |
| Priority | normal |
| Resolution | Unresolved |
| Component | Runtime System |
| Test case | |
| Differential revisions | |
| BlockedBy | |
| Related | |
| Blocking | |
| CC | |
| Operating system | |
| Architecture |
Relates to
- #92618.0.13
Activity
added RTS Tbug Trac import labels
Attached file
task-reap.patch(download).Trac metadata
Trac field Value CC → ndmitchell@gmail.com, simonmar Having to call in to dealloc has some issues:
- When destroying a thread, you need to know if it was ever used for GHC or incurr some overhead. As long as the dealloc is cheap and doesn't require you to have done the alloc, that's OK.
- You need to know when the thread goes away. If you are using a thread pool library, that's often very tricky. In our particular case it's feasible.
- If a thread dies unexpectedly the memory will be leaked. That's hopefully not a big issue in practice.
So an explicit dealloc is OK for us, with the caveat that is must work successfully on a thread that was never used by GHC.
changed milestone to %7.8.2
assigned to @simonmar
Thinking about it a bit more, it ought to be possible to clean these up automatically. The tricky bit is knowing when to do so - too eager and we penalize threads that are doing frequent in-calls, too lazy and we have the memory leak again. Perhaps just sweeping the table during a major GC would be good enough. Better might be to only free Tasks that have survived one major GC cycle without being used, but that's a bit more work.
I'll look into this when I've got some time.
Trac metadata
Trac field Value Priority normal → highest mentioned in commit af6746fb
changed milestone to %7.8.1
Simon’s commit broke the CI builds, see: https://s3.amazonaws.com/archive.travis-ci.org/jobs/19838358/log.txt
=====> T8124(normal) 2080 of 3905 [0, 0, 0] Compile failed (status 256) errors were: <command line>: does not exist: T8124_c.c *** unexpected failure for T8124(normal)mentioned in commit 3fba8759
Sorry, me again. On one of my development machines, I get
Compile failed (status 256) errors were: [1 of 1] Compiling T8124 ( T8124.hs, T8124.o ) Linking T8124 ... Warning: -rtsopts and -with-rtsopts have no effect with -no-hs-main. Call hs_init_ghc() from your main() function to set these options. /usr/bin/ld: T8124_c.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5' /lib/x86_64-linux-gnu/libpthread.so.0: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status *** unexpected failure for T8124(normal)Any ideas?
No, doesn’t help, sorry. Note that it does succeeds in the
threadedways:====> Scanning ./all.T =====> T8124(normal) 47 of 48 [0, 0, 0] cd . && $MAKE -s --no-print-directory T8124_setup cd . && '/data1/breitner/ghc/ghc-T7994/inplace/bin/ghc-stage2' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-ghci-history -o T8124 T8124.hs T8124_c.c -no-hs-main -optl-lpthread >T8124.comp.stderr 2>&1 Compile failed (status 256) errors were: [1 of 1] Compiling T8124 ( T8124.hs, T8124.o ) Linking T8124 ... Warning: -rtsopts and -with-rtsopts have no effect with -no-hs-main. Call hs_init_ghc() from your main() function to set these options. /usr/bin/ld: T8124_c.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5' /lib/x86_64-linux-gnu/libpthread.so.0: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status *** unexpected failure for T8124(normal) =====> T8124(hpc) 47 of 48 [0, 1, 0] cd . && $MAKE -s --no-print-directory T8124_setup cd . && '/data1/breitner/ghc/ghc-T7994/inplace/bin/ghc-stage2' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-ghci-history -o T8124 T8124.hs -O -fhpc -hpcdir .hpc.T8124 T8124_c.c -no-hs-main -optl-lpthread >T8124.comp.stderr 2>&1 Compile failed (status 256) errors were: [1 of 1] Compiling T8124 ( T8124.hs, T8124.o ) Linking T8124 ... Warning: -rtsopts and -with-rtsopts have no effect with -no-hs-main. Call hs_init_ghc() from your main() function to set these options. /usr/bin/ld: T8124_c.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5' /lib/x86_64-linux-gnu/libpthread.so.0: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status *** unexpected failure for T8124(hpc) =====> T8124(optasm) 47 of 48 [0, 2, 0] cd . && $MAKE -s --no-print-directory T8124_setup cd . && '/data1/breitner/ghc/ghc-T7994/inplace/bin/ghc-stage2' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-ghci-history -o T8124 T8124.hs -O -fasm T8124_c.c -no-hs-main -optl-lpthread >T8124.comp.stderr 2>&1 Compile failed (status 256) errors were: [1 of 1] Compiling T8124 ( T8124.hs, T8124.o ) Linking T8124 ... Warning: -rtsopts and -with-rtsopts have no effect with -no-hs-main. Call hs_init_ghc() from your main() function to set these options. /usr/bin/ld: T8124_c.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5' /lib/x86_64-linux-gnu/libpthread.so.0: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status *** unexpected failure for T8124(optasm) =====> T8124(threaded1) 47 of 48 [0, 3, 0] cd . && $MAKE -s --no-print-directory T8124_setup cd . && '/data1/breitner/ghc/ghc-T7994/inplace/bin/ghc-stage2' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-ghci-history -o T8124 T8124.hs -threaded -debug T8124_c.c -no-hs-main -optl-lpthread >T8124.comp.stderr 2>&1 cd . && ./T8124 </dev/null >T8124.run.stdout 2>T8124.run.stderr =====> T8124(threaded2) 47 of 48 [0, 3, 0] cd . && $MAKE -s --no-print-directory T8124_setup cd . && '/data1/breitner/ghc/ghc-T7994/inplace/bin/ghc-stage2' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-ghci-history -o T8124 T8124.hs -O -threaded -eventlog T8124_c.c -no-hs-main -optl-lpthread >T8124.comp.stderr 2>&1 cd . && ./T8124 +RTS -N2 -ls -RTS </dev/null >T8124.run.stdout 2>T8124.run.stderr =====> T8124(dyn) 47 of 48 [0, 3, 0] cd . && $MAKE -s --no-print-directory T8124_setup cd . && '/data1/breitner/ghc/ghc-T7994/inplace/bin/ghc-stage2' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-ghci-history -o T8124 T8124.hs -O -dynamic T8124_c.c -no-hs-main -optl-lpthread >T8124.comp.stderr 2>&1 Compile failed (status 256) errors were: [1 of 1] Compiling T8124 ( T8124.hs, T8124.o ) Linking T8124 ... Warning: -rtsopts and -with-rtsopts have no effect with -no-hs-main. Call hs_init_ghc() from your main() function to set these options. /usr/bin/ld: T8124_c.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5' /lib/x86_64-linux-gnu/libpthread.so.0: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status *** unexpected failure for T8124(dyn)Set ways appropriately in [0014fb3d]; please merge that as well so that we have a reliable test suite on the stable branch.
Can we get a merge of this soon? I think it would be desireable if we get useful information about the stable branch again on https://travis-ci.org/nomeata/ghc-complete/builds.
It seems to be on the list of things to merge: https://ghc.haskell.org/trac/ghc/wiki/Status/GHC-7.8/RC2.
Austin?
Simon
mentioned in issue #9261 (closed)
added runtime perf label
added Phigh label
