€•Œsphinx.addnodes”Œdocument”“”)”}”(Œ rawsource”Œ”Œchildren”]”(Œ translations”Œ LanguagesNode”“”)”}”(hhh]”(hŒ pending_xref”“”)”}”(hhh]”Œdocutils.nodes”ŒText”“”ŒChinese (Simplified)”…””}”Œparent”hsbaŒ attributes”}”(Œids”]”Œclasses”]”Œnames”]”Œdupnames”]”Œbackrefs”]”Œ refdomain”Œstd”Œreftype”Œdoc”Œ reftarget”Œ*/translations/zh_CN/locking/robust-futexes”Œmodname”NŒ classname”NŒ refexplicit”ˆuŒtagname”hhh ubh)”}”(hhh]”hŒChinese (Traditional)”…””}”hh2sbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ*/translations/zh_TW/locking/robust-futexes”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒItalian”…””}”hhFsbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ*/translations/it_IT/locking/robust-futexes”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒJapanese”…””}”hhZsbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ*/translations/ja_JP/locking/robust-futexes”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒKorean”…””}”hhnsbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ*/translations/ko_KR/locking/robust-futexes”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒSpanish”…””}”hh‚sbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ*/translations/sp_SP/locking/robust-futexes”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubeh}”(h]”h ]”h"]”h$]”h&]”Œcurrent_language”ŒEnglish”uh1h hhŒ _document”hŒsource”NŒline”NubhŒsection”“”)”}”(hhh]”(hŒtitle”“”)”}”(hŒ(A description of what robust futexes are”h]”hŒ(A description of what robust futexes are”…””}”(hh¨hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h¦hh£hžhhŸŒD/var/lib/git/docbuild/linux/Documentation/locking/robust-futexes.rst”h KubhŒ field_list”“”)”}”(hhh]”hŒfield”“”)”}”(hhh]”(hŒ field_name”“”)”}”(hŒ Started by”h]”hŒ Started by”…””}”(hhÃhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÁhh¾hŸh¶h KubhŒ field_body”“”)”}”(hŒIngo Molnar ”h]”hŒ paragraph”“”)”}”(hŒIngo Molnar ”h]”(hŒ Ingo Molnar <”…””}”(hhÙhžhhŸNh NubhŒ reference”“”)”}”(hŒmingo@redhat.com”h]”hŒmingo@redhat.com”…””}”(hhãhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”Œrefuri”Œmailto:mingo@redhat.com”uh1háhhÙubhŒ>”…””}”(hhÙhžhhŸNh Nubeh}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KhhÓubah}”(h]”h ]”h"]”h$]”h&]”uh1hÑhh¾ubeh}”(h]”h ]”h"]”h$]”h&]”uh1h¼hŸh¶h Khh¹hžhubah}”(h]”h ]”h"]”h$]”h&]”uh1h·hh£hžhhŸh¶h Kubh¢)”}”(hhh]”(h§)”}”(hŒ Background”h]”hŒ Background”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h¦hjhžhhŸh¶h KubhØ)”}”(hŒêwhat are robust futexes? To answer that, we first need to understand what futexes are: normal futexes are special types of locks that in the noncontended case can be acquired/released from userspace without having to enter the kernel.”h]”hŒêwhat are robust futexes? To answer that, we first need to understand what futexes are: normal futexes are special types of locks that in the noncontended case can be acquired/released from userspace without having to enter the kernel.”…””}”(hj hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K hjhžhubhØ)”}”(hX´A futex is in essence a user-space address, e.g. a 32-bit lock variable field. If userspace notices contention (the lock is already owned and someone else wants to grab it too) then the lock is marked with a value that says "there's a waiter pending", and the sys_futex(FUTEX_WAIT) syscall is used to wait for the other guy to release it. The kernel creates a 'futex queue' internally, so that it can later on match up the waiter with the waker - without them having to know about each other. When the owner thread releases the futex, it notices (via the variable value) that there were waiter(s) pending, and does the sys_futex(FUTEX_WAKE) syscall to wake them up. Once all waiters have taken and released the lock, the futex is again back to 'uncontended' state, and there's no in-kernel state associated with it. The kernel completely forgets that there ever was a futex at that address. This method makes futexes very lightweight and scalable.”h]”hXÄA futex is in essence a user-space address, e.g. a 32-bit lock variable field. If userspace notices contention (the lock is already owned and someone else wants to grab it too) then the lock is marked with a value that says “there’s a waiter pendingâ€, and the sys_futex(FUTEX_WAIT) syscall is used to wait for the other guy to release it. The kernel creates a ‘futex queue’ internally, so that it can later on match up the waiter with the waker - without them having to know about each other. When the owner thread releases the futex, it notices (via the variable value) that there were waiter(s) pending, and does the sys_futex(FUTEX_WAKE) syscall to wake them up. Once all waiters have taken and released the lock, the futex is again back to ‘uncontended’ state, and there’s no in-kernel state associated with it. The kernel completely forgets that there ever was a futex at that address. This method makes futexes very lightweight and scalable.”…””}”(hj.hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KhjhžhubhØ)”}”(hXi"Robustness" is about dealing with crashes while holding a lock: if a process exits prematurely while holding a pthread_mutex_t lock that is also shared with some other process (e.g. yum segfaults while holding a pthread_mutex_t, or yum is kill -9-ed), then waiters for that lock need to be notified that the last owner of the lock exited in some irregular way.”h]”hXm“Robustness†is about dealing with crashes while holding a lock: if a process exits prematurely while holding a pthread_mutex_t lock that is also shared with some other process (e.g. yum segfaults while holding a pthread_mutex_t, or yum is kill -9-ed), then waiters for that lock need to be notified that the last owner of the lock exited in some irregular way.”…””}”(hj<hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KhjhžhubhØ)”}”(hŒöTo solve such types of problems, "robust mutex" userspace APIs were created: pthread_mutex_lock() returns an error value if the owner exits prematurely - and the new owner can decide whether the data protected by the lock can be recovered safely.”h]”hŒúTo solve such types of problems, “robust mutex†userspace APIs were created: pthread_mutex_lock() returns an error value if the owner exits prematurely - and the new owner can decide whether the data protected by the lock can be recovered safely.”…””}”(hjJhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K%hjhžhubhØ)”}”(hXïThere is a big conceptual problem with futex based mutexes though: it is the kernel that destroys the owner task (e.g. due to a SEGFAULT), but the kernel cannot help with the cleanup: if there is no 'futex queue' (and in most cases there is none, futexes being fast lightweight locks) then the kernel has no information to clean up after the held lock! Userspace has no chance to clean up after the lock either - userspace is the one that crashes, so it has no opportunity to clean up. Catch-22.”h]”hXóThere is a big conceptual problem with futex based mutexes though: it is the kernel that destroys the owner task (e.g. due to a SEGFAULT), but the kernel cannot help with the cleanup: if there is no ‘futex queue’ (and in most cases there is none, futexes being fast lightweight locks) then the kernel has no information to clean up after the held lock! Userspace has no chance to clean up after the lock either - userspace is the one that crashes, so it has no opportunity to clean up. Catch-22.”…””}”(hjXhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K*hjhžhubhØ)”}”(hŒ§In practice, when e.g. yum is kill -9-ed (or segfaults), a system reboot is needed to release that futex based lock. This is one of the leading bugreports against yum.”h]”hŒ§In practice, when e.g. yum is kill -9-ed (or segfaults), a system reboot is needed to release that futex based lock. This is one of the leading bugreports against yum.”…””}”(hjfhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K2hjhžhubhØ)”}”(hXªTo solve this problem, the traditional approach was to extend the vma (virtual memory area descriptor) concept to have a notion of 'pending robust futexes attached to this area'. This approach requires 3 new syscall variants to sys_futex(): FUTEX_REGISTER, FUTEX_DEREGISTER and FUTEX_RECOVER. At do_exit() time, all vmas are searched to see whether they have a robust_head set. This approach has two fundamental problems left:”h]”hX®To solve this problem, the traditional approach was to extend the vma (virtual memory area descriptor) concept to have a notion of ‘pending robust futexes attached to this area’. This approach requires 3 new syscall variants to sys_futex(): FUTEX_REGISTER, FUTEX_DEREGISTER and FUTEX_RECOVER. At do_exit() time, all vmas are searched to see whether they have a robust_head set. This approach has two fundamental problems left:”…””}”(hjthžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K6hjhžhubhŒ block_quote”“”)”}”(hŒ×- it has quite complex locking and race scenarios. The vma-based approach had been pending for years, but they are still not completely reliable. - they have to scan _every_ vma at sys_exit() time, per thread! ”h]”hŒ bullet_list”“”)”}”(hhh]”(hŒ list_item”“”)”}”(hŒit has quite complex locking and race scenarios. The vma-based approach had been pending for years, but they are still not completely reliable. ”h]”hØ)”}”(hŒit has quite complex locking and race scenarios. The vma-based approach had been pending for years, but they are still not completely reliable.”h]”hŒit has quite complex locking and race scenarios. The vma-based approach had been pending for years, but they are still not completely reliable.”…””}”(hj“hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K>hjubah}”(h]”h ]”h"]”h$]”h&]”uh1jhjŠubjŽ)”}”(hŒ>they have to scan _every_ vma at sys_exit() time, per thread! ”h]”hØ)”}”(hŒ=they have to scan _every_ vma at sys_exit() time, per thread!”h]”hŒ=they have to scan _every_ vma at sys_exit() time, per thread!”…””}”(hj«hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KBhj§ubah}”(h]”h ]”h"]”h$]”h&]”uh1jhjŠubeh}”(h]”h ]”h"]”h$]”h&]”Œbullet”Œ-”uh1jˆhŸh¶h K>hj„ubah}”(h]”h ]”h"]”h$]”h&]”uh1j‚hŸh¶h K>hjhžhubhØ)”}”(hŒõThe second disadvantage is a real killer: pthread_exit() takes around 1 microsecond on Linux, but with thousands (or tens of thousands) of vmas every pthread_exit() takes a millisecond or more, also totally destroying the CPU's L1 and L2 caches!”h]”hŒ÷The second disadvantage is a real killer: pthread_exit() takes around 1 microsecond on Linux, but with thousands (or tens of thousands) of vmas every pthread_exit() takes a millisecond or more, also totally destroying the CPU’s L1 and L2 caches!”…””}”(hjÍhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KDhjhžhubhØ)”}”(hX‰This is very much noticeable even for normal process sys_exit_group() calls: the kernel has to do the vma scanning unconditionally! (this is because the kernel has no knowledge about how many robust futexes there are to be cleaned up, because a robust futex might have been registered in another task, and the futex variable might have been simply mmap()-ed into this process's address space).”h]”hX‹This is very much noticeable even for normal process sys_exit_group() calls: the kernel has to do the vma scanning unconditionally! (this is because the kernel has no knowledge about how many robust futexes there are to be cleaned up, because a robust futex might have been registered in another task, and the futex variable might have been simply mmap()-ed into this process’s address space).”…””}”(hjÛhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KIhjhžhubhØ)”}”(hŒÔThis huge overhead forced the creation of CONFIG_FUTEX_ROBUST so that normal kernels can turn it off, but worse than that: the overhead makes robust futexes impractical for any type of generic Linux distribution.”h]”hŒÔThis huge overhead forced the creation of CONFIG_FUTEX_ROBUST so that normal kernels can turn it off, but worse than that: the overhead makes robust futexes impractical for any type of generic Linux distribution.”…””}”(hjéhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KPhjhžhubhØ)”}”(hŒSo something had to be done.”h]”hŒSo something had to be done.”…””}”(hj÷hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KThjhžhubeh}”(h]”Œ background”ah ]”h"]”Œ background”ah$]”h&]”uh1h¡hh£hžhhŸh¶h Kubh¢)”}”(hhh]”(h§)”}”(hŒNew approach to robust futexes”h]”hŒNew approach to robust futexes”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h¦hj hžhhŸh¶h KWubhØ)”}”(hXwAt the heart of this new approach there is a per-thread private list of robust locks that userspace is holding (maintained by glibc) - which userspace list is registered with the kernel via a new syscall [this registration happens at most once per thread lifetime]. At do_exit() time, the kernel checks this user-space list: are there any robust futex locks to be cleaned up?”h]”hXwAt the heart of this new approach there is a per-thread private list of robust locks that userspace is holding (maintained by glibc) - which userspace list is registered with the kernel via a new syscall [this registration happens at most once per thread lifetime]. At do_exit() time, the kernel checks this user-space list: are there any robust futex locks to be cleaned up?”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KYhj hžhubhØ)”}”(hXüIn the common case, at do_exit() time, there is no list registered, so the cost of robust futexes is just a simple current->robust_list != NULL comparison. If the thread has registered a list, then normally the list is empty. If the thread/process crashed or terminated in some incorrect way then the list might be non-empty: in this case the kernel carefully walks the list [not trusting it], and marks all locks that are owned by this thread with the FUTEX_OWNER_DIED bit, and wakes up one waiter (if any).”h]”hXüIn the common case, at do_exit() time, there is no list registered, so the cost of robust futexes is just a simple current->robust_list != NULL comparison. If the thread has registered a list, then normally the list is empty. If the thread/process crashed or terminated in some incorrect way then the list might be non-empty: in this case the kernel carefully walks the list [not trusting it], and marks all locks that are owned by this thread with the FUTEX_OWNER_DIED bit, and wakes up one waiter (if any).”…””}”(hj,hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K`hj hžhubhØ)”}”(hŒ}The list is guaranteed to be private and per-thread at do_exit() time, so it can be accessed by the kernel in a lockless way.”h]”hŒ}The list is guaranteed to be private and per-thread at do_exit() time, so it can be accessed by the kernel in a lockless way.”…””}”(hj:hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Kihj hžhubhØ)”}”(hXdThere is one race possible though: since adding to and removing from the list is done after the futex is acquired by glibc, there is a few instructions window for the thread (or process) to die there, leaving the futex hung. To protect against this possibility, userspace (glibc) also maintains a simple per-thread 'list_op_pending' field, to allow the kernel to clean up if the thread dies after acquiring the lock, but just before it could have added itself to the list. Glibc sets this list_op_pending field before it tries to acquire the futex, and clears it after the list-add (or list-remove) has finished.”h]”hXhThere is one race possible though: since adding to and removing from the list is done after the futex is acquired by glibc, there is a few instructions window for the thread (or process) to die there, leaving the futex hung. To protect against this possibility, userspace (glibc) also maintains a simple per-thread ‘list_op_pending’ field, to allow the kernel to clean up if the thread dies after acquiring the lock, but just before it could have added itself to the list. Glibc sets this list_op_pending field before it tries to acquire the futex, and clears it after the list-add (or list-remove) has finished.”…””}”(hjHhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Klhj hžhubhØ)”}”(hŒ|That's all that is needed - all the rest of robust-futex cleanup is done in userspace [just like with the previous patches].”h]”hŒ~That’s all that is needed - all the rest of robust-futex cleanup is done in userspace [just like with the previous patches].”…””}”(hjVhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Kvhj hžhubhØ)”}”(hŒvUlrich Drepper has implemented the necessary glibc support for this new mechanism, which fully enables robust mutexes.”h]”hŒvUlrich Drepper has implemented the necessary glibc support for this new mechanism, which fully enables robust mutexes.”…””}”(hjdhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Kyhj hžhubhØ)”}”(hŒXKey differences of this userspace-list based approach, compared to the vma based method:”h]”hŒXKey differences of this userspace-list based approach, compared to the vma based method:”…””}”(hjrhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K|hj hžhubjƒ)”}”(hX- it's much, much faster: at thread exit time, there's no need to loop over every vma (!), which the VM-based method has to do. Only a very simple 'is the list empty' op is done. - no VM changes are needed - 'struct address_space' is left alone. - no registration of individual locks is needed: robust mutexes don't need any extra per-lock syscalls. Robust mutexes thus become a very lightweight primitive - so they don't force the application designer to do a hard choice between performance and robustness - robust mutexes are just as fast. - no per-lock kernel allocation happens. - no resource limits are needed. - no kernel-space recovery call (FUTEX_RECOVER) is needed. - the implementation and the locking is "obvious", and there are no interactions with the VM. ”h]”j‰)”}”(hhh]”(jŽ)”}”(hŒ±it's much, much faster: at thread exit time, there's no need to loop over every vma (!), which the VM-based method has to do. Only a very simple 'is the list empty' op is done. ”h]”hØ)”}”(hŒ°it's much, much faster: at thread exit time, there's no need to loop over every vma (!), which the VM-based method has to do. Only a very simple 'is the list empty' op is done.”h]”hŒ¸it’s much, much faster: at thread exit time, there’s no need to loop over every vma (!), which the VM-based method has to do. Only a very simple ‘is the list empty’ op is done.”…””}”(hj‹hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Khj‡ubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubjŽ)”}”(hŒAno VM changes are needed - 'struct address_space' is left alone. ”h]”hØ)”}”(hŒ@no VM changes are needed - 'struct address_space' is left alone.”h]”hŒDno VM changes are needed - ‘struct address_space’ is left alone.”…””}”(hj£hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KƒhjŸubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubjŽ)”}”(hX'no registration of individual locks is needed: robust mutexes don't need any extra per-lock syscalls. Robust mutexes thus become a very lightweight primitive - so they don't force the application designer to do a hard choice between performance and robustness - robust mutexes are just as fast. ”h]”hØ)”}”(hX&no registration of individual locks is needed: robust mutexes don't need any extra per-lock syscalls. Robust mutexes thus become a very lightweight primitive - so they don't force the application designer to do a hard choice between performance and robustness - robust mutexes are just as fast.”h]”hX*no registration of individual locks is needed: robust mutexes don’t need any extra per-lock syscalls. Robust mutexes thus become a very lightweight primitive - so they don’t force the application designer to do a hard choice between performance and robustness - robust mutexes are just as fast.”…””}”(hj»hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K…hj·ubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubjŽ)”}”(hŒ'no per-lock kernel allocation happens. ”h]”hØ)”}”(hŒ&no per-lock kernel allocation happens.”h]”hŒ&no per-lock kernel allocation happens.”…””}”(hjÓhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K‹hjÏubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubjŽ)”}”(hŒno resource limits are needed. ”h]”hØ)”}”(hŒno resource limits are needed.”h]”hŒno resource limits are needed.”…””}”(hjëhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Khjçubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubjŽ)”}”(hŒ9no kernel-space recovery call (FUTEX_RECOVER) is needed. ”h]”hØ)”}”(hŒ8no kernel-space recovery call (FUTEX_RECOVER) is needed.”h]”hŒ8no kernel-space recovery call (FUTEX_RECOVER) is needed.”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Khjÿubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubjŽ)”}”(hŒ\the implementation and the locking is "obvious", and there are no interactions with the VM. ”h]”hØ)”}”(hŒ[the implementation and the locking is "obvious", and there are no interactions with the VM.”h]”hŒ_the implementation and the locking is “obviousâ€, and there are no interactions with the VM.”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K‘hjubah}”(h]”h ]”h"]”h$]”h&]”uh1jhj„ubeh}”(h]”h ]”h"]”h$]”h&]”jÅjÆuh1jˆhŸh¶h Khj€ubah}”(h]”h ]”h"]”h$]”h&]”uh1j‚hŸh¶h Khj hžhubeh}”(h]”Œnew-approach-to-robust-futexes”ah ]”h"]”Œnew approach to robust futexes”ah$]”h&]”uh1h¡hh£hžhhŸh¶h KWubh¢)”}”(hhh]”(h§)”}”(hŒ Performance”h]”hŒ Performance”…””}”(hjFhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h¦hjChžhhŸh¶h K•ubhØ)”}”(hŒ†I have benchmarked the time needed for the kernel to process a list of 1 million (!) held locks, using the new method [on a 2GHz CPU]:”h]”hŒ†I have benchmarked the time needed for the kernel to process a list of 1 million (!) held locks, using the new method [on a 2GHz CPU]:”…””}”(hjThžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K—hjChžhubjƒ)”}”(hŒj- with FUTEX_WAIT set [contended mutex]: 130 msecs - without FUTEX_WAIT set [uncontended mutex]: 30 msecs ”h]”j‰)”}”(hhh]”(jŽ)”}”(hŒ0with FUTEX_WAIT set [contended mutex]: 130 msecs”h]”hØ)”}”(hjkh]”hŒ0with FUTEX_WAIT set [contended mutex]: 130 msecs”…””}”(hjmhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h Kšhjiubah}”(h]”h ]”h"]”h$]”h&]”uh1jhjfubjŽ)”}”(hŒ5without FUTEX_WAIT set [uncontended mutex]: 30 msecs ”h]”hØ)”}”(hŒ4without FUTEX_WAIT set [uncontended mutex]: 30 msecs”h]”hŒ4without FUTEX_WAIT set [uncontended mutex]: 30 msecs”…””}”(hj„hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K›hj€ubah}”(h]”h ]”h"]”h$]”h&]”uh1jhjfubeh}”(h]”h ]”h"]”h$]”h&]”jÅjÆuh1jˆhŸh¶h Kšhjbubah}”(h]”h ]”h"]”h$]”h&]”uh1j‚hŸh¶h KšhjChžhubhØ)”}”(hŒæI have also measured an approach where glibc does the lock notification [which it currently does for !pshared robust mutexes], and that took 256 msecs - clearly slower, due to the 1 million FUTEX_WAKE syscalls userspace had to do.”h]”hŒæI have also measured an approach where glibc does the lock notification [which it currently does for !pshared robust mutexes], and that took 256 msecs - clearly slower, due to the 1 million FUTEX_WAKE syscalls userspace had to do.”…””}”(hj¤hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KhjChžhubhØ)”}”(hŒ£(1 million held locks are unheard of - we expect at most a handful of locks to be held at a time. Nevertheless it's nice to know that this approach scales nicely.)”h]”hŒ¥(1 million held locks are unheard of - we expect at most a handful of locks to be held at a time. Nevertheless it’s nice to know that this approach scales nicely.)”…””}”(hj²hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K¢hjChžhubeh}”(h]”Œ performance”ah ]”h"]”Œ performance”ah$]”h&]”uh1h¡hh£hžhhŸh¶h K•ubh¢)”}”(hhh]”(h§)”}”(hŒImplementation details”h]”hŒImplementation details”…””}”(hjËhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h¦hjÈhžhhŸh¶h K§ubhØ)”}”(hŒsThe patch adds two new syscalls: one to register the userspace list, and one to query the registered list pointer::”h]”hŒrThe patch adds two new syscalls: one to register the userspace list, and one to query the registered list pointer:”…””}”(hjÙhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K©hjÈhžhubhŒ literal_block”“”)”}”(hŒðasmlinkage long sys_set_robust_list(struct robust_list_head __user *head, size_t len); asmlinkage long sys_get_robust_list(int pid, struct robust_list_head __user **head_ptr, size_t __user *len_ptr);”h]”hŒðasmlinkage long sys_set_robust_list(struct robust_list_head __user *head, size_t len); asmlinkage long sys_get_robust_list(int pid, struct robust_list_head __user **head_ptr, size_t __user *len_ptr);”…””}”hjésbah}”(h]”h ]”h"]”h$]”h&]”Œ xml:space”Œpreserve”uh1jçhŸh¶h K¬hjÈhžhubhØ)”}”(hXList registration is very fast: the pointer is simply stored in current->robust_list. [Note that in the future, if robust futexes become widespread, we could extend sys_clone() to register a robust-list head for new threads, without the need of another syscall.]”h]”hXList registration is very fast: the pointer is simply stored in current->robust_list. [Note that in the future, if robust futexes become widespread, we could extend sys_clone() to register a robust-list head for new threads, without the need of another syscall.]”…””}”(hjùhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K´hjÈhžhubhØ)”}”(hX8So there is virtually zero overhead for tasks not using robust futexes, and even for robust futex users, there is only one extra syscall per thread lifetime, and the cleanup operation, if it happens, is fast and straightforward. The kernel doesn't have any internal distinction between robust and normal futexes.”h]”hX:So there is virtually zero overhead for tasks not using robust futexes, and even for robust futex users, there is only one extra syscall per thread lifetime, and the cleanup operation, if it happens, is fast and straightforward. The kernel doesn’t have any internal distinction between robust and normal futexes.”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K¹hjÈhžhubhØ)”}”(hŒbIf a futex is found to be held at exit time, the kernel sets the following bit of the futex word::”h]”hŒaIf a futex is found to be held at exit time, the kernel sets the following bit of the futex word:”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h K¿hjÈhžhubjè)”}”(hŒ*#define FUTEX_OWNER_DIED 0x40000000”h]”hŒ*#define FUTEX_OWNER_DIED 0x40000000”…””}”hj#sbah}”(h]”h ]”h"]”h$]”h&]”j÷jøuh1jçhŸh¶h KÂhjÈhžhubhØ)”}”(hŒUand wakes up the next futex waiter (if any). User-space does the rest of the cleanup.”h]”hŒUand wakes up the next futex waiter (if any). User-space does the rest of the cleanup.”…””}”(hj1hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÄhjÈhžhubhØ)”}”(hŒ‡Otherwise, robust futexes are acquired by glibc by putting the TID into the futex field atomically. Waiters set the FUTEX_WAITERS bit::”h]”hŒ†Otherwise, robust futexes are acquired by glibc by putting the TID into the futex field atomically. Waiters set the FUTEX_WAITERS bit:”…””}”(hj?hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÇhjÈhžhubjè)”}”(hŒ*#define FUTEX_WAITERS 0x80000000”h]”hŒ*#define FUTEX_WAITERS 0x80000000”…””}”hjMsbah}”(h]”h ]”h"]”h$]”h&]”j÷jøuh1jçhŸh¶h KÊhjÈhžhubhØ)”}”(hŒ'and the remaining bits are for the TID.”h]”hŒ'and the remaining bits are for the TID.”…””}”(hj[hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÌhjÈhžhubeh}”(h]”Œimplementation-details”ah ]”h"]”Œimplementation details”ah$]”h&]”uh1h¡hh£hžhhŸh¶h K§ubh¢)”}”(hhh]”(h§)”}”(hŒTesting, architecture support”h]”hŒTesting, architecture support”…””}”(hjthžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h¦hjqhžhhŸh¶h KÏubhØ)”}”(hŒ¢I've tested the new syscalls on x86 and x86_64, and have made sure the parsing of the userspace list is robust [ ;-) ] even if the list is deliberately corrupted.”h]”hŒ¤I’ve tested the new syscalls on x86 and x86_64, and have made sure the parsing of the userspace list is robust [ ;-) ] even if the list is deliberately corrupted.”…””}”(hj‚hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÑhjqhžhubhØ)”}”(hŒ i386 and x86_64 syscalls are wired up at the moment, and Ulrich has tested the new glibc code (on x86_64 and i386), and it works for his robust-mutex testcases.”h]”hŒ i386 and x86_64 syscalls are wired up at the moment, and Ulrich has tested the new glibc code (on x86_64 and i386), and it works for his robust-mutex testcases.”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÕhjqhžhubhØ)”}”(hŒ^All other architectures should build just fine too - but they won't have the new syscalls yet.”h]”hŒ`All other architectures should build just fine too - but they won’t have the new syscalls yet.”…””}”(hjžhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÙhjqhžhubhØ)”}”(hŒwArchitectures need to implement the new futex_atomic_cmpxchg_inatomic() inline function before writing up the syscalls.”h]”hŒwArchitectures need to implement the new futex_atomic_cmpxchg_inatomic() inline function before writing up the syscalls.”…””}”(hj¬hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1h×hŸh¶h KÜhjqhžhubeh}”(h]”Œtesting-architecture-support”ah ]”h"]”Œtesting, architecture support”ah$]”h&]”uh1h¡hh£hžhhŸh¶h KÏubeh}”(h]”Œ(a-description-of-what-robust-futexes-are”ah ]”h"]”Œ(a description of what robust futexes are”ah$]”h&]”uh1h¡hhhžhhŸh¶h Kubeh}”(h]”h ]”h"]”h$]”h&]”Œsource”h¶uh1hŒcurrent_source”NŒ current_line”NŒsettings”Œdocutils.frontend”ŒValues”“”)”}”(h¦NŒ generator”NŒ datestamp”NŒ source_link”NŒ source_url”NŒ toc_backlinks”Œentry”Œfootnote_backlinks”KŒ sectnum_xform”KŒstrip_comments”NŒstrip_elements_with_classes”NŒ strip_classes”NŒ report_level”KŒ halt_level”KŒexit_status_level”KŒdebug”NŒwarning_stream”NŒ traceback”ˆŒinput_encoding”Œ utf-8-sig”Œinput_encoding_error_handler”Œstrict”Œoutput_encoding”Œutf-8”Œoutput_encoding_error_handler”jíŒerror_encoding”Œutf-8”Œerror_encoding_error_handler”Œbackslashreplace”Œ language_code”Œen”Œrecord_dependencies”NŒconfig”NŒ id_prefix”hŒauto_id_prefix”Œid”Œ dump_settings”NŒdump_internals”NŒdump_transforms”NŒdump_pseudo_xml”NŒexpose_internals”NŒstrict_visitor”NŒ_disable_config”NŒ_source”h¶Œ _destination”NŒ _config_files”]”Œ7/var/lib/git/docbuild/linux/Documentation/docutils.conf”aŒfile_insertion_enabled”ˆŒ raw_enabled”KŒline_length_limit”M'Œpep_references”NŒ pep_base_url”Œhttps://peps.python.org/”Œpep_file_url_template”Œpep-%04d”Œrfc_references”NŒ rfc_base_url”Œ&https://datatracker.ietf.org/doc/html/”Œ tab_width”KŒtrim_footnote_reference_space”‰Œsyntax_highlight”Œlong”Œ smart_quotes”ˆŒsmartquotes_locales”]”Œcharacter_level_inline_markup”‰Œdoctitle_xform”‰Œ docinfo_xform”KŒsectsubtitle_xform”‰Œ image_loading”Œlink”Œembed_stylesheet”‰Œcloak_email_addresses”ˆŒsection_self_link”‰Œenv”NubŒreporter”NŒindirect_targets”]”Œsubstitution_defs”}”Œsubstitution_names”}”Œrefnames”}”Œrefids”}”Œnameids”}”(jÇjÄj jj@j=jÅjÂjnjkj¿j¼uŒ nametypes”}”(jljj ‰j@‰jʼnjn‰j¿‰uh}”(jÄh£jjj=j jÂjCjkjÈj¼jquŒ footnote_refs”}”Œ citation_refs”}”Œ autofootnotes”]”Œautofootnote_refs”]”Œsymbol_footnotes”]”Œsymbol_footnote_refs”]”Œ footnotes”]”Œ citations”]”Œautofootnote_start”KŒsymbol_footnote_start”KŒ id_counter”Œ collections”ŒCounter”“”}”…”R”Œparse_messages”]”Œtransform_messages”]”Œ transformer”NŒ include_log”]”Œ decoration”Nhžhub.