TODO revision 1.1.2.16
1 1.1.2.9 nathanw Bugs to fix: 2 1.1.2.9 nathanw 3 1.1.2.9 nathanw - pthread_cond_timedwait() doesn't work if SA's aren't running yet, 4 1.1.2.9 nathanw because the alarm system isn't up and running. It would be weird to 5 1.1.2.9 nathanw use them that way, but it's perfectly legal. 6 1.1.2.9 nathanw - There is a race between pthread_cancel() and 7 1.1.2.9 nathanw pthread_cond_broadcast() or pthread_exit() about removing an item 8 1.1.2.9 nathanw from the sleep queue. The locking protocols there need a little 9 1.1.2.9 nathanw adjustment. 10 1.1.2.11 thorpej - pthread_sig.c: pthread__signal_tramp() is broken. It gets a ucontext, 11 1.1.2.11 thorpej but then hands off an empty (garbage from stack!) sigcontext to the 12 1.1.2.11 thorpej handler. The ucontext MUST be converted to a sigcontext, the handler 13 1.1.2.11 thorpej called, and then the sigcontext converted back to ucontext, before the 14 1.1.2.11 thorpej ucontext is used to perform the sigreturn-analogue. This is necessary 15 1.1.2.11 thorpej for e.g. C++/Java exceptions and some garbage-collectors. 16 1.1.2.11 thorpej - pthread_sig.c: pthread__signal_tramp() should be changed so that it is 17 1.1.2.11 thorpej easy to add support for SA_SIGINFO later. (Mostly a function signature 18 1.1.2.11 thorpej issue.) 19 1.1.2.15 thorpej - pthread_sig.c: Come up with a signal trampoline naming convention like 20 1.1.2.15 thorpej libc's, so that GDB will have an easier time with things. 21 1.1.2.11 thorpej - Consider moving pthread__signal_tramp() to its own file, and building 22 1.1.2.11 thorpej it with -fasync-unwind-tables, so that DWARF2 EH unwinding works through 23 1.1.2.11 thorpej it. (This is required for e.g. GCC's libjava.) 24 1.1.2.11 thorpej - locks: Add support for RAS on uniprocessors. 25 1.1.2.11 thorpej - pthread_lock.c: pthread_spin_lock() should use a loop around 26 1.1.2.11 thorpej __cpu_simple_lock_try() so that it is easy to add support for RAS. 27 1.1.2.12 thorpej - Add locking to ld.elf_so so that multiple threads doing lazy binding 28 1.1.2.12 thorpej doesn't trash things. 29 1.1.2.13 thorpej - Verify the cancel stub symbol trickery. 30 1.1.2.9 nathanw 31 1.1.2.1 nathanw 32 1.1.2.3 nathanw Interfaces/features to implement: 33 1.1.2.3 nathanw - pthread_kill() 34 1.1.2.3 nathanw - pthread_atfork() 35 1.1.2.3 nathanw - priority scheduling 36 1.1.2.3 nathanw - libc integration: 37 1.1.2.3 nathanw - foo_r interfaces 38 1.1.2.3 nathanw - system integration 39 1.1.2.3 nathanw - some macros and prototypes belong in headers other than pthread.h 40 1.1.2.3 nathanw 41 1.1.2.3 nathanw 42 1.1.2.3 nathanw Features that need more/better regression tests: 43 1.1.2.3 nathanw - pthread_cond_broadcast() 44 1.1.2.3 nathanw - pthread_once() 45 1.1.2.3 nathanw - pthread_get/setspecific() 46 1.1.2.3 nathanw - signals 47 1.1.2.3 nathanw 48 1.1.2.3 nathanw 49 1.1.2.3 nathanw Things that need fixing: 50 1.1.2.3 nathanw - Recycle dead threads for new threads. 51 1.1.2.3 nathanw 52 1.1.2.3 nathanw Ideas to play with: 53 1.1.2.3 nathanw - Explore the trapcontext vs. usercontext distinction in ucontext_t. 54 1.1.2.3 nathanw - Get rid of thread structures when too many accumulate (is this 55 1.1.2.3 nathanw actually a good idea?) 56 1.1.2.3 nathanw - Adaptive spin/sleep locks for mutexes. 57 1.1.2.1 nathanw - Supporting different mutex types would be nice (normal, debugging, 58 1.1.2.1 nathanw recursive, etc). 59 1.1.2.1 nathanw - Currently, each thread uses two real pages of memory: one at the top 60 1.1.2.1 nathanw of the stack for actual stack data, and one at the bottom for the 61 1.1.2.1 nathanw pthread_st. If we can get suitable space above the initial stack for 62 1.1.2.1 nathanw main(), we can cut this to one page per thread. Perhaps crt0 should 63 1.1.2.1 nathanw do something different (give us more space) if libpthread is linked 64 1.1.2.1 nathanw in? 65 1.1.2.1 nathanw - Figure out whether/how to expose the inline version of 66 1.1.2.1 nathanw pthread_self(). 67
Indexes created Mon Aug 03 00:25:36 UTC 2026