[lng-odp] [Bug 1313] Timeout user pointer metadata is not set on timer_cancel
https://bugs.linaro.org/show_bug.cgi?id=1313 --- Comment #3 from Ola Liljedahl ola.liljed...@linaro.org --- http://patches.opendataplane.org/patch/1034/ v2 changes the wording of some comments. -- You are receiving this mail because: You are on the CC list for the bug.___ lng-odp mailing list lng-odp@lists.linaro.org http://lists.linaro.org/mailman/listinfo/lng-odp
[lng-odp] [Bug 1313] Timeout user pointer metadata is not set on timer_cancel
https://bugs.linaro.org/show_bug.cgi?id=1313 Maxim Uvarov maxim.uva...@linaro.org changed: What|Removed |Added Status|CONFIRMED |RESOLVED CC||maxim.uva...@linaro.org Resolution|--- |FIXED --- Comment #4 from Maxim Uvarov maxim.uva...@linaro.org --- commit dd48199df5e1d55c3290ce01747edc596e35bd4f Author: Ola Liljedahl ola.liljed...@linaro.org Date: Wed Mar 11 16:36:45 2015 +0100 linux-generic: odp_timer: set user_ptr for cancelled timeout Ensure that the timeout user_ptr and timer fields are set when the corresponding timer is immediately cancelled. https://bugs.linaro.org/show_bug.cgi?id=1313 Signed-off-by: Ola Liljedahl ola.liljed...@linaro.org Reviewed-by: Bill Fischofer bill.fischo...@linaro.org Signed-off-by: Maxim Uvarov maxim.uva...@linaro.org -- You are receiving this mail because: You are on the CC list for the bug.___ lng-odp mailing list lng-odp@lists.linaro.org http://lists.linaro.org/mailman/listinfo/lng-odp
[lng-odp] [Bug 1313] Timeout user pointer metadata is not set on timer_cancel
https://bugs.linaro.org/show_bug.cgi?id=1313 --- Comment #2 from Ola Liljedahl ola.liljed...@linaro.org --- http://patches.opendataplane.org/patch/995/ -- You are receiving this mail because: You are on the CC list for the bug.___ lng-odp mailing list lng-odp@lists.linaro.org http://lists.linaro.org/mailman/listinfo/lng-odp
[lng-odp] [Bug 1313] Timeout user pointer metadata is not set on timer_cancel
https://bugs.linaro.org/show_bug.cgi?id=1313 --- Comment #1 from Ola Liljedahl ola.liljed...@linaro.org --- The idea here was to access the timeout event as seldom as possible in order to avoid potential cache misses for data structures that are otherwise not accessed. The timeout event is (currently) not accessed when the timer is set or cancelled (which are supposed to be frequent operations). The timeout header is written (timer, expiration tick, user pointer) when the timer expires (which is supposed to be an infrequent operation). So not updating the user pointer as not an accident, merely an attempt at optimization. If the timer API semantics requires that the timeout event after cancellation is updated in a similar way to when the timer expires (at least user pointer has to be valid), of course the implementation has to be changed. I think the doxygen documentation needs be updated as well, it is not obvious what are the requirements for timeouts from cancelled timers. I can see two approaches. 1) Update the timeout header in odp_timer_cancel() the same way it is updated when the timer expires. 2) Always update the timeout header when the timer is set and don't update the timeout when the timer expires or is cancelled. I am looking at the implementation in order to see which choice makes best sense. To the user, the design chosen will not be observable. What should odp_timeout_tick() return for a timeout from a cancelled timer? The originally requested expiration tick or some value that indicates cancellation? Currently the value is undefined for cancelled timeouts. -- You are receiving this mail because: You are on the CC list for the bug.___ lng-odp mailing list lng-odp@lists.linaro.org http://lists.linaro.org/mailman/listinfo/lng-odp