2016-07-25T05:24:55 *** sebhub has joined #rtems 2016-07-25T05:49:53 *** monstr has joined #rtems 2016-07-25T06:36:59 *** dspace has joined #rtems 2016-07-25T06:40:43 [trac] #2752/enhancement (new) updated by Sebastian Huber (In [changeset:"0fd6f25507fbea5f4892b71b58837cdda17856b4/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2752 2016-07-25T06:40:44 [git] 3ae199a by Sebastian Huber: [score: Disable RTEMS_NO_RETURN for RTEMS_DEBUG Do not use RTEMS_NO_RETURN hints for debug configurations to ease use of stack traces in case of fatal errors.] http://devel.rtems.org/changeset/3ae199a/rtems 2016-07-25T06:40:44 [git] 0fd6f25 by Sebastian Huber: [score: Relax thread begin extension environment Update #2752.] http://devel.rtems.org/changeset/0fd6f25/rtems 2016-07-25T07:04:44 *** tasslehoff has joined #rtems 2016-07-25T07:23:30 *** mumptai has joined #rtems 2016-07-25T11:10:21 *** gigetoo has quit IRC 2016-07-25T11:58:12 *** benjamint1 has joined #rtems 2016-07-25T12:00:24 *** benjamint has quit IRC 2016-07-25T12:02:17 *** benjamint1 has quit IRC 2016-07-25T12:02:26 *** tasslehoff has quit IRC 2016-07-25T12:03:15 [git] ba5de75 by Sebastian Huber: [bsps: Fix shared polled console fatal error Fatal errors must uniquely identify the source location.] http://devel.rtems.org/changeset/ba5de75/rtems 2016-07-25T12:17:27 [trac] #2765/defect (v:4.10) created by sebastian.huber (Application level deadlocks may lead to SMP lock level deadlocks) http://devel.rtems.org/ticket/2765 2016-07-25T12:18:48 [trac] #2765/defect (new) updated by Sebastian Huber (In [changeset:"57a00bc6afd25f5c41006b386627d087ff9d4c66/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2765 2016-07-25T12:18:49 [git] 57a00bc by Sebastian Huber: [smptests/smpmutex02: New test Update #2765.] http://devel.rtems.org/changeset/57a00bc/rtems 2016-07-25T12:20:41 [trac] #2765/defect (new) updated by sebastian.huber (empty comment) http://devel.rtems.org/ticket/2765 2016-07-25T12:33:12 *** benjamint has joined #rtems 2016-07-25T12:39:43 *** benjamint has quit IRC 2016-07-25T12:48:44 [git] 7f616dd by Sebastian Huber: [score: Fix for RTEMS_DEBUG] http://devel.rtems.org/changeset/7f616dd/rtems 2016-07-25T13:02:58 *** jkridner has joined #rtems 2016-07-25T14:31:15 *** gedare has joined #rtems 2016-07-25T14:31:15 *** ChanServ sets mode: +o gedare 2016-07-25T14:31:35 *** benjamint has joined #rtems 2016-07-25T14:40:58 *** mric has joined #rtems 2016-07-25T15:10:15 *** dspace has quit IRC 2016-07-25T15:54:29 [trac] #2766/defect (v:4.10) created by dcb (/src/lib/libbsp/arm/stm32f4x/hal/STM32F4xx_HAL_Driver/Src/stm32f4xx_hal_fmpi2c.c:2891: code and comment don't match ?) http://devel.rtems.org/ticket/2766 2016-07-25T16:26:51 *** benjamint has quit IRC 2016-07-25T16:27:18 *** dr__house|away is now known as dr__house 2016-07-25T16:34:48 *** jkridner has quit IRC 2016-07-25T16:35:27 *** jkridner has joined #rtems 2016-07-25T16:39:44 *** jkridner has quit IRC 2016-07-25T16:45:22 [trac] #2732/defect (new) updated by Gedare Bloom (In [changeset:"e0f17fcf6f06d3383cd389d809ee4e3d6e0fb14d/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2732 2016-07-25T16:45:22 [git] f23d470 by Gedare Bloom: [cpukit: Add and use Watchdog_Discipline. Clock disciplines may be WATCHDOG_RELATIVE, WATCHDOG_ABSOLUTE, or WATCHDOG_NO_TIMEOUT. A discipline of WATCHDOG_RELATIVE with a timeout of WATCHDOG_NO_TIMEOUT is equivalent to a discipline of WATCHDOG_NO_TIMEOUT. updates #2732] http://devel.rtems.org/changeset/f23d470/rtems 2016-07-25T16:45:23 [git] 6e98ea9 by Gedare Bloom: [posix: add clock_nanosleep and tests updates #2732] http://devel.rtems.org/changeset/6e98ea9/rtems 2016-07-25T16:45:23 [git] 10e4000 by Gedare Bloom: [cpukit/rtems: fix return type mismatch for _TOD_To_seconds] http://devel.rtems.org/changeset/10e4000/rtems 2016-07-25T16:45:23 [git] 127c20e by Gedare Bloom: [posix: refactor cond wait support to defer abstime conversion updates #2745] http://devel.rtems.org/changeset/127c20e/rtems 2016-07-25T16:45:23 [git] b5bfaaf by Gedare Bloom: [posix: cond_timedwait remember and use clock from condattr updates #2745] http://devel.rtems.org/changeset/b5bfaaf/rtems 2016-07-25T16:45:23 [git] e0f17fc by Gedare Bloom: [posix: fix clock_nanosleep and nanosleep clock use Sleeping with CLOCK_REALTIME should use the WATCHDOG_ABSOLUTE clock discipline for the threadq so that the timeout interval may change in case the clock source changes. Similarly, CLOCK_MONOTONIC uses the WATCHDOG_RELATIVE threadq that will only wakeup the thread after the 2016-07-25T16:45:23 requested count of ticks elapse...] http://devel.rtems.org/changeset/e0f17fc/rtems 2016-07-25T17:01:42 *** itazzz is now known as ita 2016-07-25T17:46:13 *** jkridner has joined #rtems 2016-07-25T17:51:18 *** jkridner has quit IRC 2016-07-25T18:44:41 *** Davidbrcz has joined #rtems 2016-07-25T18:58:34 *** Davidbrcz has quit IRC 2016-07-25T19:12:52 *** Davidbrcz has joined #rtems 2016-07-25T19:14:08 *** dr__house is now known as dr__house|away 2016-07-25T19:18:47 *** monstr has quit IRC 2016-07-25T19:36:28 *** Wonjun_ has joined #rtems 2016-07-25T19:39:33 *** Wonjun__ has quit IRC 2016-07-25T20:16:53 *** dr__house|away is now known as dr__house 2016-07-25T20:53:15 *** gedare has quit IRC 2016-07-25T21:41:13 *** stryx` has joined #rtems 2016-07-25T22:14:27 *** jkridner has joined #rtems 2016-07-25T22:24:35 *** Davidbrcz has quit IRC 2016-07-25T22:37:04 *** mumptai has quit IRC 2016-07-25T23:03:26 *** gigetoo has joined #rtems 2016-07-26T00:16:46 *** dr__house is now known as dr__house|away 2016-07-26T03:28:22 [git] f78a5f5 by Chris Johns: [score: Add a STATES_DEBUGGER for use by debugging agents to a thread's state.] http://devel.rtems.org/changeset/f78a5f5/rtems 2016-07-26T04:30:40 *** Wonjun_ has quit IRC 2016-07-26T04:42:11 *** Wonjun has joined #rtems 2016-07-26T05:33:22 *** monstr has joined #rtems 2016-07-26T05:54:38 [git] 00c620c by Sebastian Huber: [score: Format] http://devel.rtems.org/changeset/00c620c/rtems 2016-07-26T05:54:38 [git] 75fc27a by Sebastian Huber: [score: Fix printk() format specifiers] http://devel.rtems.org/changeset/75fc27a/rtems 2016-07-26T06:02:22 [git] ea93844 by Sebastian Huber: [score: Format] http://devel.rtems.org/changeset/ea93844/rtems 2016-07-26T06:22:11 *** mumptai has joined #rtems 2016-07-26T07:57:34 [git] c1b815a by Sebastian Huber: [sptests/spextensions01: Avoid NULL pointer access] http://devel.rtems.org/changeset/c1b815a/rtems 2016-07-26T08:03:53 [git] 1a8fe67 by Alexander Krutwig: [Add Untar_FromChunk_Print() + Test] http://devel.rtems.org/changeset/1a8fe67/rtems 2016-07-26T08:04:03 [git] 6a174c0 by Alexander Krutwig: [Add Untar_FromGzChunk_Print() + Test] http://devel.rtems.org/changeset/6a174c0/rtems 2016-07-26T08:04:12 [git] 84008f9 by Alexander Krutwig: [bsps/arm: Clear pending exceptions for ARMv7-M] http://devel.rtems.org/changeset/84008f9/rtems 2016-07-26T08:04:22 [git] c51e6d9 by Alexander Krutwig: [atsam: Use normal memory for DTCM Avoids problems with unaligned data access.] http://devel.rtems.org/changeset/c51e6d9/rtems 2016-07-26T08:20:16 *** benjamint has joined #rtems 2016-07-26T08:23:17 sebhub: hi 2016-07-26T08:33:52 sebhub: see email 2016-07-26T09:16:40 *** mumptai has quit IRC 2016-07-26T09:51:46 *** Wonjun has quit IRC 2016-07-26T10:35:24 *** kiwichri_ has joined #rtems 2016-07-26T10:35:43 sebhub, hi 2016-07-26T11:08:02 hi chris 2016-07-26T11:08:20 in which time zone are you currently? 2016-07-26T11:22:19 Sydney 2016-07-26T11:22:37 then is quite late/early? 2016-07-26T11:22:44 It is getting late. 2016-07-26T11:24:42 I would like to find an agreeable solution to reusing the low level code for debugging support. I am not sure what is best but I using the rtems_fatal error path is not right, it might be possible but I think we need to find a better solution. 2016-07-26T11:25:00 It seems a waste of time and effort duplicating it. 2016-07-26T11:25:45 this stuff is very target dependent 2016-07-26T11:26:06 Yes and just like CPU_Exception_frame we can define something working. 2016-07-26T11:26:14 workable 2016-07-26T11:26:40 you can add a jump to CPU_Exception_frame per target 2016-07-26T11:26:58 and invoke this in the fatal error handler 2016-07-26T11:27:08 is speed really important for single step debugging? 2016-07-26T11:27:17 I do not want the fatal error path, it is not a fatal error 2016-07-26T11:27:30 the fatal path is recoverable via a jump 2016-07-26T11:27:36 It is either the wrong name or the wrong way to handle it. 2016-07-26T11:28:02 yes in range stepping it is 2016-07-26T11:28:17 This is where a range is given and you step in that range 2016-07-26T11:29:13 I don't think its really slow on ARM if you use the fatal error and a jump 2016-07-26T11:29:45 I did not check the User extension path. 2016-07-26T11:30:00 Is the user extension installable at runtime? 2016-07-26T11:30:03 if you need something fast, then you need a dedicated exception handler via arm_cp15_set_exception_handler() 2016-07-26T11:30:10 yes 2016-07-26T11:30:26 Yes this is a good idea but it ends up needing aligned memory etc etc 2016-07-26T11:30:47 aligned memory for what? 2016-07-26T11:31:33 I seem to remember the vector table needs to aligned but I forget to what size and then I need to duplicate the asm code. 2016-07-26T11:32:19 the table is already there 2016-07-26T11:32:25 you just replace a table entry 2016-07-26T11:32:36 c/src/lib/libbsp/arm/shared/arm-gic-irq.c 2016-07-26T11:32:38 for example 2016-07-26T11:33:07 OK so that is asm wrapper. 2016-07-26T11:33:19 What is _ARM_Exception_default is made weak? 2016-07-26T11:33:23 What if _ARM_Exception_default is made weak? 2016-07-26T11:34:10 there is not return path, so this _ARM_Exception_default must be a no return function 2016-07-26T11:34:21 Ah yes of course 2016-07-26T11:34:31 you may need also some mode changes 2016-07-26T11:34:43 OK, this means arm-gic-irq.c 2016-07-26T11:35:04 mode changes? 2016-07-26T11:37:43 On the idea of a handler pointer like i386 and ARM use to have, the corruption can still happen but at the USER handler level unless you are thinking about const init tables. 2016-07-26T11:44:08 do you mean thumb/arm mode changes? 2016-07-26T11:53:27 no, the armv4 has supervisor and data abort mode for example 2016-07-26T11:54:41 the modes have some shadow registers 2016-07-26T12:00:46 Ah ok. 2016-07-26T12:00:56 I am off, thanks. 2016-07-26T12:01:28 *** kiwichri_ has quit IRC 2016-07-26T12:18:31 [git] f6fdf2e by Alexander Krutwig: [atsam: Add user defined RX/TX descriptor counts] http://devel.rtems.org/changeset/f6fdf2e/rtems 2016-07-26T12:40:17 *** tplk has joined #rtems 2016-07-26T13:27:35 *** gedare has joined #rtems 2016-07-26T13:27:35 *** ChanServ sets mode: +o gedare 2016-07-26T13:45:54 *** gedare_ has joined #rtems 2016-07-26T13:45:54 *** gedare has quit IRC 2016-07-26T13:57:47 *** gedare_ has quit IRC 2016-07-26T13:58:00 *** gedare_ has joined #rtems 2016-07-26T13:58:00 *** gedare_ is now known as gedare 2016-07-26T13:58:27 *** gedare has joined #rtems 2016-07-26T13:58:27 *** ChanServ sets mode: +o gedare 2016-07-26T14:42:20 *** dougk_ has joined #rtems 2016-07-26T15:46:03 *** monstr has quit IRC 2016-07-26T16:21:45 *** tplk has quit IRC 2016-07-26T17:17:33 *** dr__house|away is now known as dr__house 2016-07-26T17:18:02 *** monstr has joined #rtems 2016-07-26T17:28:11 *** mumptai has joined #rtems 2016-07-26T18:17:57 [trac] #2732/defect (new) updated by Gedare Bloom (In [changeset:"39d97ab78cfcc37eb8b1e7d9f9717f51b249155b/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2732 2016-07-26T18:17:58 [git] 39d97ab by Gedare Bloom: [cpukit: refactor nanosleep and use 64-bit timeout for threadq * Fixes a bug with elapsed time calculations misusing absolute time arguments in nanosleep_helper by passing the requested relative interval. * Fixes a bug with truncation of absolute timeouts by passing the full 64-bit value to Thread_queue_Enqueue. * Share yield 2016-07-26T18:17:58 logic between nanosleep a...] http://devel.rtems.org/changeset/39d97ab/rtems 2016-07-26T18:39:06 *** monstr has quit IRC 2016-07-26T20:02:09 *** Davidbrcz has joined #rtems 2016-07-26T20:50:43 *** gedare has quit IRC 2016-07-26T20:53:48 *** dr__house is now known as dr__house|away 2016-07-26T21:29:52 *** Davidbrcz has quit IRC 2016-07-26T22:27:49 *** mumptai has quit IRC 2016-07-26T22:39:40 [trac] GSoC/2016 edited by Sambeet (empty comment) http://devel.rtems.org/wiki/GSoC/2016 2016-07-27T04:47:35 *** monstr has joined #rtems 2016-07-27T05:36:38 *** mumptai has joined #rtems 2016-07-27T06:35:53 sebhub: hi. 2016-07-27T06:36:07 sebhub: it looks a change to _Thread_queue_Enqueue_critical has broken libbsd. 2016-07-27T06:36:27 ../../rtemsbsd/rtems/rtems-kernel-muteximpl.c:64:3: error: too many arguments to function '_Thread_queue_Enqueue_critical' 2016-07-27T06:37:41 is this one of Gedare's changes? 2016-07-27T06:41:22 .yes 2016-07-27T06:41:54 Thanks, I will post something. 2016-07-27T06:42:26 do you fix it or should I do it? 2016-07-27T06:42:41 Can you please do it? 2016-07-27T06:42:54 It looks simple but I have not looked at this code in detail. 2016-07-27T06:43:04 yes, I was about to do it 2016-07-27T06:43:13 Thanks 2016-07-27T06:52:23 [git] e5aa554 by Sebastian Huber: [rtems-bsd-mutex: Update due to API changes] http://devel.rtems.org/changeset/e5aa554/rtems-libbsd 2016-07-27T06:53:38 sebhub: thank you, I have a complete from source for tools, rtems and libbsd on Windows10. 2016-07-27T06:53:50 complete build that is 2016-07-27T06:54:14 all on windows, no linux to windows cross compiler involved? 2016-07-27T06:54:20 yes 2016-07-27T06:54:28 nice, on cygwin? 2016-07-27T06:54:32 MSYS2 2016-07-27T06:54:38 and yes it is nice. 2016-07-27T06:54:48 The user manual documents the procedure 2016-07-27T06:55:38 how much time needs the build of the tools? 2016-07-27T06:56:06 hard to tell cause I am on a VM and I have limited the cores to 2 and memory to 4 or 8G 2016-07-27T06:56:28 more than one day or less? 2016-07-27T06:56:34 I will try a fast i7 cad machines one day soon 2016-07-27T06:57:14 Build Set: Time 1:46:34.429180 2016-07-27T06:57:17 for the tools 2016-07-27T06:57:47 some years a go I tried to do it on cygwin, it took more than one day to build the powerpc compiler 2016-07-27T06:58:27 MSYS2 is really nice and pacman is neat. You get a real git, ssh plus more so no need to have extra bits installed 2016-07-27T06:58:39 I have tmux running 2016-07-27T07:10:07 [git] dfa2cdb by Sebastian Huber: [posix: Fix error status] http://devel.rtems.org/changeset/dfa2cdb/rtems 2016-07-27T07:49:27 *** dougk_ has quit IRC 2016-07-27T08:06:27 *** mumptai has quit IRC 2016-07-27T08:11:10 *** dougk_ has joined #rtems 2016-07-27T08:14:51 *** dougk_ has quit IRC 2016-07-27T08:27:12 *** dougk_ has joined #rtems 2016-07-27T08:56:07 [git] 0b83633 by Sebastian Huber: [score: Fix for RTEMS_DEBUG] http://devel.rtems.org/changeset/0b83633/rtems 2016-07-27T08:56:16 [trac] #2556/enhancement (accepted) updated by Sebastian Huber (In [changeset:"f4d1f307926b6319e5d3b325dbe424901285dec1/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2556 2016-07-27T08:56:16 [trac] #2412/enhancement (assigned) updated by Sebastian Huber (In [changeset:"f4d1f307926b6319e5d3b325dbe424901285dec1/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2412 2016-07-27T08:56:16 [trac] #2765/defect (new) updated by Sebastian Huber (In [changeset:"f4d1f307926b6319e5d3b325dbe424901285dec1/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2765 2016-07-27T08:56:17 [git] f4d1f30 by Sebastian Huber: [score: Split _Thread_Change_priority() Update #2412. Update #2556. Update #2765.] http://devel.rtems.org/changeset/f4d1f30/rtems 2016-07-27T08:56:26 [trac] #2556/enhancement (accepted) updated by Sebastian Huber (In [changeset:"ac8402ddd6e4a8eb6defb98220d39d4c20a6f025/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2556 2016-07-27T08:56:26 [trac] #2412/enhancement (assigned) updated by Sebastian Huber (In [changeset:"ac8402ddd6e4a8eb6defb98220d39d4c20a6f025/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2412 2016-07-27T08:56:27 [trac] #2765/defect (new) updated by Sebastian Huber (In [changeset:"ac8402ddd6e4a8eb6defb98220d39d4c20a6f025/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2765 2016-07-27T08:56:27 [git] ac8402d by Sebastian Huber: [score: Simplify _Thread_queue_Boost_priority() Raise the priority under thread queue lock protection and omit the superfluous thread queue priority change, since the thread is extracted anyway. The unblock operation will pick up the new priority. Update #2412. Update #2556. Update #2765.] 2016-07-27T08:56:27 http://devel.rtems.org/changeset/ac8402d/rtems 2016-07-27T08:56:36 [trac] #2556/enhancement (accepted) updated by Sebastian Huber (In [changeset:"3a58dc863157bb21054a144c1a21b690544c0d23/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2556 2016-07-27T08:56:36 [trac] #2412/enhancement (assigned) updated by Sebastian Huber (In [changeset:"3a58dc863157bb21054a144c1a21b690544c0d23/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2412 2016-07-27T08:56:37 [trac] #2765/defect (new) updated by Sebastian Huber (In [changeset:"3a58dc863157bb21054a144c1a21b690544c0d23/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2765 2016-07-27T08:56:37 [git] 3a58dc8 by Sebastian Huber: [score: Priority inherit thread queue operations Move the priority change due to priority interitance to the thread queue enqueue operation to simplify the locking on SMP configurations. Update #2412. Update #2556. Update #2765.] http://devel.rtems.org/changeset/3a58dc8/rtems 2016-07-27T08:56:47 [trac] #2556/enhancement (accepted) updated by Sebastian Huber (In [changeset:"1fcac5adc5ed38fb88ce4c6d24b2ca2e27e3cd10/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2556 2016-07-27T08:56:47 [trac] #2412/enhancement (assigned) updated by Sebastian Huber (In [changeset:"1fcac5adc5ed38fb88ce4c6d24b2ca2e27e3cd10/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2412 2016-07-27T08:56:47 [trac] #2765/defect (new) updated by Sebastian Huber (In [changeset:"1fcac5adc5ed38fb88ce4c6d24b2ca2e27e3cd10/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2765 2016-07-27T08:56:48 [git] 1fcac5a by Sebastian Huber: [score: Turn thread lock into thread wait lock The _Thread_Lock_acquire() function had a potentially infinite run-time due to the lack of fairness at atomic operations level. Update #2412. Update #2556. Update #2765.] http://devel.rtems.org/changeset/1fcac5a/rtems 2016-07-27T08:56:57 [trac] #2556/enhancement (accepted) updated by Sebastian Huber (In [changeset:"d79df38c2bea50112214ade95776cb90d693e390/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2556 2016-07-27T08:56:57 [trac] #2412/enhancement (assigned) updated by Sebastian Huber (In [changeset:"d79df38c2bea50112214ade95776cb90d693e390/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2412 2016-07-27T08:56:58 [trac] #2765/defect (closed) updated by Sebastian Huber (In [changeset:"d79df38c2bea50112214ade95776cb90d693e390/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2765 2016-07-27T08:56:58 [git] d79df38 by Sebastian Huber: [score: Add deadlock detection The mutex objects use the owner field of the thread queues for the mutex owner. Use this and add a deadlock detection to _Thread_queue_Enqueue_critical() for thread queues with an owner. Update #2412. Update #2556. Close #2765.] http://devel.rtems.org/changeset/d79df38/rtems 2016-07-27T10:09:00 *** seadog007 has joined #rtems 2016-07-27T13:39:12 *** deval has joined #rtems 2016-07-27T13:56:47 *** salpha has joined #rtems 2016-07-27T13:58:12 *** punitvara has joined #rtems 2016-07-27T13:58:59 *** gedare has joined #rtems 2016-07-27T13:58:59 *** ChanServ sets mode: +o gedare 2016-07-27T13:59:36 howdy 2016-07-27T13:59:54 Hello 2016-07-27T14:00:24 *** Wonjun has joined #rtems 2016-07-27T14:00:32 aloha 2016-07-27T14:01:04 Good morning. 2016-07-27T14:01:12 *** DrJoel has joined #rtems 2016-07-27T14:01:13 *** DrJoel has joined #rtems 2016-07-27T14:01:13 *** ChanServ sets mode: +o DrJoel 2016-07-27T14:01:25 let's get started, deval you may go first 2016-07-27T14:01:44 Hi 2016-07-27T14:01:45 I have completed adding USB support and primitive Ethernet support on Raspberry Pi. As of now I am getting Ethernet address "ue0: Ethernet address: 5a:ee:60:74:67:92", and name of the USB devices attached (if the related drivers are in the place). 2016-07-27T14:01:51 "umass1: on usbus0" this message was for a USB pendrive I have. Since the "umass" driver was there, I got the message. 2016-07-27T14:01:58 Now, as per my appication I have to add HID devices like Keyboard and mouse. For that I will have to add drivers related to them. On which I am working currently and looks like a quite time taking process. https://lists.rtems.org/pipermail/devel/2016-July/015672.html 2016-07-27T14:02:09 I have a question here. Should I continue adding the drivers for HIDs and complete the USB porting or devote my reaining time adding FreeBSD TCP/IP stack followed by lwIP ? 2016-07-27T14:03:03 Have you used a network application yet? 2016-07-27T14:03:25 No Not yet 2016-07-27T14:03:41 Personal opinion ... network stack is more important and probably close. Followed closely by testing mass storage. 2016-07-27T14:04:29 i agree. work on the networking stack, and get your code merged too 2016-07-27T14:04:38 +1 on merging. 2016-07-27T14:05:02 Even I have the same opinion, but I had to ask here first. 2016-07-27T14:05:08 The ping and ftpd01 tests are good to focus on. 2016-07-27T14:05:09 okay 2016-07-27T14:05:16 I will do that then 2016-07-27T14:05:46 thanks deval, salpha you'll be next. anything else deval? 2016-07-27T14:05:49 Nearly 100% of Pi RTEMS users will want a TCP/IP stack. Few will really want graphics 2016-07-27T14:06:02 Nothing from my side . 2016-07-27T14:06:06 Thank you :) 2016-07-27T14:06:07 OK 2016-07-27T14:06:14 i updated build instructions and minor edits to vagrant scripts 2016-07-27T14:06:46 work should be re-producable from dev branch now 2016-07-27T14:07:00 i faced some issues while developing apps using libdl 2016-07-27T14:07:11 and we have a discussion on the mailing list 2016-07-27T14:07:26 now i want to try waf instructions that Chris posted 2016-07-27T14:07:44 *** mudit1729 has joined #rtems 2016-07-27T14:07:50 i also thought about "deliverables" 2016-07-27T14:08:14 does it make sense to inject crashed test cases into the already-developed rtems apps to see how they behave? 2016-07-27T14:08:51 by already-developed apps i mean more serious apps that are potentially deployed somewhere 2016-07-27T14:09:19 it makes some sense, but I'm not sure what the value is exactly 2016-07-27T14:10:12 due to changes to the program state, test cases that already have failed might not fail in another scenario (program) 2016-07-27T14:10:25 or test cases that haven't fail, might fail somewhere else 2016-07-27T14:10:48 oh i see. 2016-07-27T14:11:10 or i don't know. do you have ideas about "deliverables"? 2016-07-27T14:11:30 what is most important to the community that you proposed this project? 2016-07-27T14:11:35 I really just mean that your code is polished and documentation is available 2016-07-27T14:12:06 that you write a good report too :) 2016-07-27T14:12:18 yeah but this will happen probably in the third and fourth week of Sep during the wrap-up 2016-07-27T14:12:34 I think you will have trouble getting access to a "real app" and being able to run it on random HW. The best option off the top of my head would be Alan Cudmore's Core Flight System demo. 2016-07-27T14:12:49 But having deliverables is first priority. 2016-07-27T14:12:59 It would be interesting to see if Alan is interested. 2016-07-27T14:13:02 okay we'll talk about the report. i think almost everything is there in the blog. maybe i need to re-structure it though 2016-07-27T14:13:39 ok, and make your code in good shape, proper licenses, etc 2016-07-27T14:13:43 i haven't had a chance to look through it yet 2016-07-27T14:13:54 thanks salpha, Wonjun you'll be next 2016-07-27T14:13:56 okay. then let's talk to Alan after i finished changing test execution to the dynamic model 2016-07-27T14:14:18 probably around end of August or earlier, if i find some time in-between my exams 2016-07-27T14:14:39 ok thank you. Wonjun are you here? 2016-07-27T14:14:51 yes 2016-07-27T14:14:53 thank you too. plz have a look and give me feedbacks so that i know what matters to you more. 2016-07-27T14:15:05 ok, go ahead Wonjun. 2016-07-27T14:15:11 Hello, Until now, I was working on writing a config file I have worked. 2016-07-27T14:15:27 last week, I couldn't start rtems cell. but I started rtems as cell without any error but also no outputs. 2016-07-27T14:15:27 Moreover I found If I execute rtems example Jan sent,my computer have e100 network driver. 2016-07-27T14:15:36 I will ask Jan about the reason why I had no output when I started rtems and how to check what happend. 2016-07-27T14:15:51 And I have some mails I asked from mentos 2016-07-27T14:16:03 There are lots of issues about my situations. 2016-07-27T14:16:21 but I am little bit confused... 2016-07-27T14:17:07 ok. so one thing that I'm still not satisfied with is the application of the patches Jan previously sent. I would have liked if you kept each patch as a separate commit, and added whatever commits you needed to merge them correctly 2016-07-27T14:17:20 instead of what you have, which is a single squashed commit of all the patches he sent you 2016-07-27T14:17:39 yes 2016-07-27T14:17:39 The patches need to be in small reviewable units. 2016-07-27T14:17:55 I am confused. Can you run an RTEMS application in the container now? 2016-07-27T14:17:58 Hello world? 2016-07-27T14:18:00 Ticker? 2016-07-27T14:18:02 etc 2016-07-27T14:18:19 Rtems Jan patched? 2016-07-27T14:18:22 or 2016-07-27T14:18:27 master? 2016-07-27T14:18:36 for now you can continue to work in the pc386 bsp directly, but it may be necessary later to create an independent "jailhouse" BSP for the i386. 2016-07-27T14:18:49 Jan's and your patches in small units to be reviewable and mergable on the master 2016-07-27T14:18:53 it might also be easier to just make your own jailhouse BSP. 2016-07-27T14:19:01 ok I can run master rtems application. 2016-07-27T14:19:59 could you give me a advice? there are various parts of patches like network pci ioacpi etc. 2016-07-27T14:20:29 you ought to be able to ignore network and ioapic i suspect 2016-07-27T14:20:35 to get a simple working version 2016-07-27T14:20:52 I'm sorry to interrupt, would it be possible for me to go next 2016-07-27T14:22:06 Wonjun: I agree with gedare that a separate BSP is likely simpler and cleaner. Focus on small patches that can be merged against the head. Be able to run hello, then ticker, then the network stack 2016-07-27T14:22:28 yes mudit1729 you can even go now. Wonjun we can try to talk a bit more after this meeting 2016-07-27T14:22:33 Thanks. I will send a mail about simple working version in detail. 2016-07-27T14:22:40 Wonjun: good idea. 2016-07-27T14:22:50 Thanks. 2016-07-27T14:22:59 So ya, I have added code for DMA support in the Mainline for RTEMS. It is available on my github repo. 2016-07-27T14:23:04 https://github.com/spark1729/rtems/commit/f3c7b6e5aec903dedc893a9b0db819ccd021a333 2016-07-27T14:23:18 eedless to say that the library is been added to the build process and builds fine. 2016-07-27T14:23:27 *Needless 2016-07-27T14:23:38 It needs to be posted to devel@, reviewed and merged. :) 2016-07-27T14:23:40 Few points to note : The api presently provided will be - rpi_dma_allocate(int req_ch), rpi_dma_reset(int ch); rpi_dma_free(int ch); rpi_dma_setup_src(int ch, int dreq, int inc_addr, int width); rpi_dma_setup_dst(int ch, int dreq, int inc_addr, int width); rpi_dma_start(int ch, vm_paddr_t src, vm_paddr_t dst, int len); rpi_dma_init(); rpi_dma_length(int ch); 2016-07-27T14:23:45 What drivers use it? 2016-07-27T14:24:08 I will be adding DMA access for SPI and I2C devices 2016-07-27T14:24:44 The basic framework is as follows : There is a control block structure that is used to control the channels. The address of which is written into a register. Important thing to note is that the DMA controller is directly connected to the peripherals, so it deals with real address spaces. In addition to that, cache coherency is also required when you consider changing addresses from virtual to real or vice versa. All these have 2016-07-27T14:25:06 The software context is basically a array of structures which entails information regarding the various channels. I have a doubt with the particular portion of the dma_init function, where it maps the control block buffer into the bus address space. 2016-07-27T14:25:17 basically what DrJoel said, you should get code review. was this code provided by sebastian huber, or did you write it? The copyright block indicates embedded brains on it. 2016-07-27T14:26:12 I wrote it from ground up, however I did borrow the primary logic from the FreeBSD DMA driver 2016-07-27T14:26:15 the commit message should be a little bit more descriptive, and doesn't need to say it is for the "RTEMS Mainline"--we know that part. 2016-07-27T14:26:29 Is there any jobs to deal with additionally? 2016-07-27T14:27:23 Wonjun: not that i can think of, try to find out why you might not get output (does the inmate have access to use COM1?), and start to work on a simpler approach 2016-07-27T14:27:57 Thanks. 2016-07-27T14:28:11 mudit1729: you should also be aware of https://devel.rtems.org/wiki/Developer/Coding/Conventions 2016-07-27T14:28:21 i notice there are for example some tabs in your code 2016-07-27T14:28:47 Yes, I will be running uncrustify on the files before I submit the patch 2016-07-27T14:28:53 ok 2016-07-27T14:28:58 I require inputs from the community 2016-07-27T14:29:12 regarding a particular portion in the dma_init function 2016-07-27T14:29:13 you'll get the best inputs by submitting the code to devel mailnig list 2016-07-27T14:29:27 or in by asking a pointed question about the same 2016-07-27T14:29:49 okay. Ill polish it a bit and submit it on the mailing list. 2016-07-27T14:30:06 you can send an email if you have some particular doubts about your code too 2016-07-27T14:30:23 okay. 2016-07-27T14:31:05 thanks mudit1729. 2016-07-27T14:31:10 punitvara: you can go next, hi beng-nl 2016-07-27T14:31:16 I will update the mailing list with the code as well as a writeup regarding the same 2016-07-27T14:31:18 There are many more people on devel@ and you have much better chances of getting useful feedback 2016-07-27T14:31:42 I am still working on I2C coding part. Till now I have tested register API of I2C 2016-07-27T14:31:57 One more thing I wanted to ask. 2016-07-27T14:32:26 Right now I am developing I/O APIs . First focus it to develop read API for bus driver 2016-07-27T14:32:46 I have added the code in the irq controller for SD card controller. Still there is some issue with the same. From a deliverable perspective, can I expand the functionality of Mailbox further as an alternative for the SD card driver 2016-07-27T14:33:04 Then Next week I will probably finish testing with Read API with reading from EEPROM 2016-07-27T14:33:11 punitvara: have you asked about APIs you might copy from? 2016-07-27T14:33:15 I will again be trying to add the SD card driver in a cleaner fashion, however I did want to ask this question 2016-07-27T14:33:45 *** ppisa has joined #rtems 2016-07-27T14:34:09 mudit1729: Is there a BSD driver for the SD card? 2016-07-27T14:34:09 mudit1729: you may if it will not overly distract you from your efforts to get the SD driver to work. otherwise you should focus on figuring out what is going wrong with the driver 2016-07-27T14:34:46 Last week I have asked on ml about reference for dev/framework. Sebastian suggested to refer cadance and atsam i2c driver. 2016-07-27T14:34:53 Joel : Yes there is. 2016-07-27T14:35:01 Gedare : Okay will look into the same. 2016-07-27T14:35:07 Thanks for your time folks 2016-07-27T14:35:40 Ok. Reuse what exists as much as possible. To a large extent, these System on Chips are just a bunch of reusable peripheral logic thrown together. 2016-07-27T14:35:50 so I am writing code from scratch for I2C. However I tried to copy the code I can from those two drivers. 2016-07-27T14:36:43 Drjoel: Yeah atsam and cadence has similar kind of logic. 2016-07-27T14:37:13 I am reusing as much as possible. 2016-07-27T14:37:51 ok thanks punitvara anything else? 2016-07-27T14:37:53 Any suggestion for testing with EEPROM ? I think I have to write slave EEPROM driver right ? 2016-07-27T14:38:25 I have seen libtest/i2c . Seems that it can be helpful for testing 2016-07-27T14:38:37 In the past, I have just written a program that open/writes/closes then opens/reads/closes. Use various patterns and reuse the write/read test logic 2016-07-27T14:39:51 darnir: are you here? 2016-07-27T14:39:52 https://github.com/punitvara/rtems/blob/new/testsuites/samples/i2c0/init.c 2016-07-27T14:40:18 I was able to test this successfully I think you mean similar kind of test 2016-07-27T14:40:41 yeah punitvara that is sufficient direction for testing 2016-07-27T14:40:55 ok thanks :) 2016-07-27T14:42:53 Everyone.. ask questions on devel@. Don't struggle. Also post code for review and merger as soon as it is ready. Otherwise, you all will swamp the core team 2016-07-27T14:42:57 ok i think that is everyone who is here... 2016-07-27T14:43:16 (Don't struggle "needlessly") ;) 2016-07-27T14:43:33 +1 no shame in asking for help :) 2016-07-27T14:43:51 yeah Thanks @Drjoel and @Gedare for your precious time :) 2016-07-27T14:44:09 ok good mtg. Hey DrJoel. I'm re-working nanosleep again. Sebastian pointed out to me that probably the use of absolute time isn't right for it. 2016-07-27T14:44:38 sure punitvara we do it with the hopes you'll save us time someday later. 2016-07-27T14:44:50 (not a sound investment strategy, but it pays off now and then.) 2016-07-27T14:45:10 I would like to discuss future direction for time base, but that is not for GSoC hour probably. 2016-07-27T14:45:12 http://pubs.opengroup.org/onlinepubs/9699919799/functions/clock_nanosleep.html says nanosleep uses CLOCK_REALTIME 2016-07-27T14:45:24 ppisa: we can chat about it now 2016-07-27T14:45:29 the gsoc hour is done now 2016-07-27T14:45:45 yeah DrJoel but is silent about what to do if the wall time is adjusted 2016-07-27T14:45:59 *** salpha has quit IRC 2016-07-27T14:46:21 OK, so if you have time, I would like to speak about my opinion and correct it by yours. 2016-07-27T14:46:25 for this reason they added the TIMER_ABSTIME attribute to the clock_nanosleep. 2016-07-27T14:46:41 sure. i'm working on it now some more any way, there seems to be some latent issues hiding still. 2016-07-27T14:46:48 turns out, time is hard. :) 2016-07-27T14:46:58 +1 2016-07-27T14:47:45 [trac] GSoC/2016 edited by hwj1593 (empty comment) http://devel.rtems.org/wiki/GSoC/2016 2016-07-27T14:48:15 From nanosleep: But, except for the case of being interrupted by a signal, the suspension time shall not be less than the time specified by rqtp, as measured by the system clock CLOCK_REALTIME. 2016-07-27T14:48:54 I know you have some concern about the clock source in the score, where we don't see an explicit handling of CLOCK_REALTIME and CLOCK_MONOTONIC. my approach was to convert all monotonic timeouts to use the internal "relative" timer, which is a monotonic clock source (not adjustable) 2016-07-27T14:49:14 In my thinking.. If you move time backwards, a nanosleep should not fire until the adjusted time shows enough has past. 2016-07-27T14:49:30 OK, my strong opinion is, that there has to be two timequeues (one "realtime/CLOCK_REALTIME" one "relative/CLOCK_MONOTONIC"). I have not so strong opinion, if they has to be per CPU or not. But if scalability to masive SMP is intended, then they has to be per CPU. 2016-07-27T14:49:52 T1 = clock_gettime(); calculate delay ; nanosleep(delay); T2 = clock_gettime() ; assert(T2 - T1 > delay) 2016-07-27T14:50:15 right, so now the "realtime" is the WATCHDOG_ABSOLUTE, and the "relative" is the WATCHDOG_RELATIVE 2016-07-27T14:50:28 Yes 2016-07-27T14:50:31 but, i think, there are only 2 system-wide. 2016-07-27T14:50:34 ppisa: For POSIX, I agree. 2016-07-27T14:52:06 There can be question for other mechanims, in Linux system it is called timer wheel, which has lower cost of deletion of timeouts which are supposed to not be reached often. Because precise queues has to be based on some real sortin which is costy for insert and even for premature removals. 2016-07-27T14:52:27 But we can left that out for now. 2016-07-27T14:53:05 DrJoel: as sebhub pointed out to me, the nanosleep doesn't say anything about what to do with wall clock adjustments. The "suspension time" as measured by the clock can be measured in relative clock ticks, although your code would then not work, huh 2016-07-27T14:53:20 BSD has wheels also but I didn't see that as a great data srtucture for RTEMS purposes. 2016-07-27T14:53:56 http://pubs.opengroup.org/onlinepubs/9699919799/functions/nanosleep.html has "But, except for the case of being interrupted by a signal, the suspension time shall not be less than the time specified by rqtp, as measured by the system clock CLOCK_REALTIME." 2016-07-27T14:54:03 which I think is what I tested in my code fragment. 2016-07-27T14:54:04 Even for "hard-real" timers there is worth to add parameter to the low level API, which allows to specify what is requested precission, tollerated firing earlier for example, for given timer call. 2016-07-27T14:54:58 Basicly QoS. Have been discussing that on another project. Unfortunately, those have to be reflected through the user API or they aren't of much value. 2016-07-27T14:55:16 *** Wonjun has quit IRC 2016-07-27T14:55:59 brb 2016-07-27T14:57:13 There is explicitly defined, that nanosleep is measured against CLOCK_REALTIME 2016-07-27T14:57:13 http://pubs.opengroup.org/onlinepubs/009695399/functions/nanosleep.html 2016-07-27T14:57:13 This means that clock adjustion should affect the interval. 2016-07-27T14:57:50 ok. that is how i have interpreted it. Sebhub asked me why. I will bring this to devel ml. 2016-07-27T14:58:22 i have it implemented now as using the WATCHDOG_ABSOLUTE that will allow for clock adjustment. 2016-07-27T14:58:31 The most uncertain think to decide is how to manage conversions to be fast/have acceptable overhead. 2016-07-27T14:58:50 yes, this is a problem 2016-07-27T14:59:25 Even Linux kernel people experimented with this multiple times and there has been different solution for 32-bit archs and 64-bit archs. 2016-07-27T15:02:17 If we do not need to convert expire time back to user time representation (usually case of timers, gettimeofday is yet another story), then conversion of all expire times (in timeval, timespec etc) to 64-bit nanoseconds is in the worst case 2x 32x32->64-bit multiply and 64-bit addition, which is not so big problem. 2016-07-27T15:03:04 So I would incline to have both WATCHDOGs in the same base unit, nanoseconds. 2016-07-27T15:03:29 that should be discussed in single proposal on devel ml 2016-07-27T15:04:01 And linear, not compacted timespec, because it makes relative times computation and conversion to absolute expire more complex. 2016-07-27T15:04:11 i did find it very confusing to deal with the multiple base units, and two different formats for "Ticks". 2016-07-27T15:05:17 well, the absolute expire uses the compacted timespec directly, so i don't understand what you mean about a conversion problem there. 2016-07-27T15:05:34 or you mean from ns->ts->absticks 2016-07-27T15:07:49 As for the ticks, it is natural for old day systems to simple the things by Ticks for the core. But modern systems tend to allows NO_HZ or other named tick-less mode. Other problem is, that in the moder systems there are multiple sources of the time. If they are based on the clock rate then tick has some meaning. But if the clocks are not synchronized between cores or you use different source with interrupt capability for times and other simpl 2016-07-27T15:07:49 time source then the value of simplification by use of clocks is mitigated. 2016-07-27T15:08:55 My opinion is that compacted timespec is contraproductive optimization. But that is for discussion. 2016-07-27T15:09:23 ok 2016-07-27T15:09:55 i would tend to agree, but the argument has to be made very clearly and simply in a single email thread for discussion 2016-07-27T15:10:25 i don't even think there are many uses of the absolute watchdog at the moment 2016-07-27T15:10:38 We should think about efectivity of conversion time source to 64-bit ns time now. 2016-07-27T15:10:58 ok. well that certainly should simplify all the internal logic 2016-07-27T15:11:53 i mean, time logic. it complicates somewhat the handling of "ticks" since multiple ns may elapse, or partial ns 2016-07-27T15:12:14 i haven't been able to study how linux/freebsd handle these concerns. 2016-07-27T15:12:38 If we consider compatibility with actual BSPs then we have some number of ns per tick. If it is integral multiple then it is easy to update 64-bit ns time (I use ns64 from now) 2016-07-27T15:13:30 The current Linux kernel does not care about ticks at all in time keeping part the fact. 2016-07-27T15:14:49 At one point, the internal time was in 64 bit nanoseconds from POSIX epoch. That made the time math code very easy to write and the representation covered a LONG time. Something like 500 years if I remember correctly 2016-07-27T15:15:00 If you have better source then periodic timer then you define mask how many bits are valid from that source. The value is converted to ns64 by simple 64-bit x 32-bit multiply and computed right shift. 2016-07-27T15:15:41 But the clock tick having the option to update by a tick and internally add a configured number of nsecs. Or the BSP to provide a variable number of nanoseconds is quite powerful. 2016-07-27T15:16:08 The Deos/RTEMS timing works better this way as does what others call tickless systems. 2016-07-27T15:16:31 FWIW I hate the term tickless. That implies a complete lack of clock ticks. It is really a variable clock tick. 2016-07-27T15:18:13 no_hz appears to be about avoiding scheduler invocations? 2016-07-27T15:18:36 Because such arithmentic is not cyclic in the respect of overflow, the valid bits mask and requirement for precission (minimal number of bits after the fisrt one) are converted to the longest time source maintenance interval. Then the 32-bit coefficient is computed and if given time source is selected as active time source then it automatically inserts and renew maintenance timer which ensures that source is read at least once before overflow 2016-07-27T15:19:58 Does any of this matter on systems which tend to have periodic threads at frequencies which don't begin to approach the range of the timer? Even a 50 millisecond period is a small length of time relative to a 32-bit timer. 2016-07-27T15:20:28 Ultimately, if the system has frequent time events, you have to have timer actions frequent enough to meet the deadlines. 2016-07-27T15:21:13 dcc chat gedare 2016-07-27T15:21:54 Tickless means that there is no periodic timer/interval in timeq queue (that does not correspond to the hardware wher even for tick system usually single shot timers with finar IRQ are requested). All times are computed on base of real need, that is timer to catch timer sourec oveflow event, timer to catch expire of current budget of task scheduled on given CPU etc. 2016-07-27T15:23:31 Do you have confidence in BSP authors to correctly reprogram the timer hardware frequently and not drift? With no external time references? 2016-07-27T15:24:07 Linux fights with need for periodic system state updateds for years. But quite some time ago has been reached state where there is no periodic timer setup for case that there is no task on given CPU runqueue, then it has been enhanced that no periodic activity is required if there is only one task on given CPU runqueues and complete full tickless mode is worked and at least proposed now. 2016-07-27T15:26:48 i see the merits. 2016-07-27T15:26:49 The precission of timmer used to fire IRQ has no integral effect in the case that you use other source for time (TSC or other source with faster access). The one schot is computed as difference between the earlier not fired expire time and actual time in ns64, than that interval is converted to the clocks of one shot timer. 2016-07-27T15:27:48 I see the merits also but I remain skeptical since I know how hard it is to get a periodic timer correct. This puts more burden on the timer device driver author and validating its correctness. Especially if it has to go through a safety audit 2016-07-27T15:28:24 they can still use a periodic clock source though, if necessary 2016-07-27T15:28:33 And each conversion and reprogramming has dead time which must be accounted for. That is where I don't have confidence that it will be easy to get right. 2016-07-27T15:29:02 Not arguing the technical side of can it work. I am worried you will get a lot of fielded systems with really odd clock skewing 2016-07-27T15:29:51 If you have only single timer hardware, then only really safe and natural way is to lefdt it run till overflow and setuap comparator each time as sum of actual value and that converted interval. It can be fater for simple HW to compute that all in the resolution of that single timer, but in Linux case which count with complex system and even runtime option to switch time sources and timer events devices independentrly is such optimization cont 2016-07-27T15:31:29 To Joel: for small RTOS and certification this can be true. 2016-07-27T15:32:49 I am not opposed to pursuing it. Just want to make sure that we go in fully aware of the target application domains of RTEMS and that we don't assume a solution from any other OS is a natural fit. 2016-07-27T15:35:34 But at the moment when you admit that there do not exists any fully trusted clock source on whole Earth and you know that your crystal oscillator has orders of magnitude lower trust than other sources and you want to network the things, then you need to tune conversion from your clock source to the MONOTONIC and REALTIME. And that fine tuning can realized at the end by that tuning of that single 64x32 shr conversion. 2016-07-27T15:36:39 Oh I admit that. And I assume you admit that every time SW touches the timer, there is overhead which must be accounted for or the time reference (no matter how good or bad it is) can shift due to that overhead not being 100% properly accounted for. 2016-07-27T15:36:48 Ignoring programming errors. :) 2016-07-27T15:36:48 ok, i think it is a reasonable direction to push toward. i don't entirely understand the arithmetic yet, but might be easier if i see the code ;) 2016-07-27T15:37:08 To Joel: safety is problem. And yes, small system with fixed cyclic schedule without preemption and withou malloc etc is hard requirement in some areas. 2016-07-27T15:38:27 I think it is a good thing to push on because it certainly can help in low power deployments. 2016-07-27T15:38:34 I need to be somewhere else in 20 minutes 2016-07-27T15:38:35 cya 2016-07-27T15:38:38 *** DrJoel has quit IRC 2016-07-27T15:39:02 *** monstr has quit IRC 2016-07-27T15:40:02 So that is a problem what is the target. The question is, do we like to have ability for more precise user timers API than "tick" is. 2016-07-27T15:41:27 Problem with tick resolution is that as application requests higher resolution then you have to use faster tick but that causes higher overhead and destroys determinism as ticks interrupt interrupts more times running code. 2016-07-27T15:42:09 there are also some clear wins for SMP systems 2016-07-27T15:42:12 On the other hand, if "user"/application can request too short intervals it can lead to DOS state. 2016-07-27T15:42:26 and low power embedded (with or without hard deadlines) 2016-07-27T15:42:32 *** punitvara has quit IRC 2016-07-27T15:42:44 might be good ways to identify those cases 2016-07-27T15:43:49 so i guess the first step would be to convert to ns64 internal timekeeping? 2016-07-27T15:44:21 If you can put together a proposal as email to devel ml we can try to gauge the developer "buy-in" for this idea. 2016-07-27T15:45:19 It worth to add API which allows to request only expire time as range and define what is the smallest range for given architecture. The APIs which do not expose precission to the user (for example nanosleep) can be treated as range 2016-07-27T15:46:23 ok 2016-07-27T15:49:05 Unfortunatelly, my time situation is worse ans worse, so I do not know when/if I have chance to push that. I have critically deffered many my tasks/duties to unblock cache and other stuff blocking RPi GSoCs. 2016-07-27T15:49:05 On the other hand this core change is area which I would use my time for. But I have to work on projects which allows to survive company and are required for school projects. 2016-07-27T15:49:43 yes of course. i would prefer to get some discussion and consensus first, then all may help as they are able. 2016-07-27T15:50:16 i have only a little more time left before new semester and i will be swamped until next summer again. 2016-07-27T15:52:22 I would like to do time sychronization right way. My friend worked on that clock fussion theory for ESA. We both have same feeling how adjustemnst should be done in the system code. But we think that even new Linux foundation effort for Linux does not go best direction. But that would need some ESA grant or something like that to can allocate man-months for that. 2016-07-27T15:53:50 seems possible if you can find the right contact who thinks this is important for RTEMS 2016-07-27T15:53:55 Yes my holidays are almost in the half now and I have only something like 5 days to refresh last month. 2016-07-27T15:54:25 if it shows a big benefit for something like leon4 it seems possible 2016-07-27T15:55:02 ah. i took some long vacations, and am getting ready for new school year. i have to write one proposal for money next week, and then one week more until classes. 2016-07-27T15:56:16 Yes, but my trider record is failure and my colleague has no luck with they whole company to get any ESA project this year as well. Even that he has knowledge and results in many areas from laser, optics and tried many proposals. 2016-07-27T15:57:26 s/trider/trader/ 2016-07-27T15:57:42 :( 2016-07-27T16:01:23 I would like to get TMS570 to complete BSP state. I have resolved unfortunate bug in serial, so it seems that I am on the right way. I have vene that minimal init code based on HalCoGen. We have XCP loader (the must for automotive use) tested and allmost integrated to our CCS codebase - woking well with older CCS support but has to be updated to actual lwIP and CCS. This would be nice environment with RTEMS application loaded over ETHERNET and 2016-07-27T16:01:23 Matlab/Simulink. We have all knowledge and allmost all done, but ... 2016-07-27T16:02:42 But I believe we push things forward one day ... so I do not need worry that there is no interesting tast for me ... 2016-07-27T16:03:08 By the way, have you already HW from Ti? 2016-07-27T16:03:24 no, thanks for reminding me, i need to remind them. donations sometimes come slow, sometimes fast. 2016-07-27T16:03:55 although, my department office moved so I don't even know the shipping address anymore. I should get that right ;) 2016-07-27T16:05:22 I wish you success. Unfortuantelly, I have not switch on our TMS570LC4357 yet. 2016-07-27T16:06:53 We move to the new building this autumn as well. We move ven labs to another part of campus in different town quarter. So lots of fun ahead. 2016-07-27T16:09:19 hope it goes well 2016-07-27T16:09:38 my lab space is scheduled for remodel sometime soon, don't know where i'll be after that. 2016-07-27T16:09:50 Anyway, I am happy that you started that clock_nanosleep move. It would be great if there is found consensus about future direction. If you have time/chance, it would worth to discuss that with Sebastian. If he agrees, then switch to linear 64-bit ns for both timers is the nect step from my point of view. 2016-07-27T16:11:03 ok 2016-07-27T16:12:52 I hope that it simplifies things and opens futire options without need to change part related to user API. 2016-07-27T16:25:37 *** benjamint has quit IRC 2016-07-27T16:28:15 Bye 2016-07-27T16:28:26 *** ppisa has left #rtems 2016-07-27T16:28:29 *** benjamint has joined #rtems 2016-07-27T16:37:18 *** deval has quit IRC 2016-07-27T17:25:43 *** mumptai has joined #rtems 2016-07-27T17:39:49 *** dr__house|away is now known as dr__house 2016-07-27T18:15:50 *** Davidbrcz has joined #rtems 2016-07-27T19:43:06 *** seadog007 has quit IRC 2016-07-27T20:24:34 *** ita is now known as itazzzz 2016-07-27T20:52:14 *** gedare has quit IRC 2016-07-27T21:55:14 *** Davidbrcz has quit IRC 2016-07-27T21:57:14 *** dr__house is now known as dr__house|away 2016-07-27T22:09:56 *** Davidbrcz has joined #rtems 2016-07-27T22:30:13 *** jkridner has quit IRC 2016-07-27T22:30:57 *** jkridner has joined #rtems 2016-07-27T22:42:06 *** Davidbrcz has quit IRC 2016-07-28T02:16:22 [git] 1bf6529 by Chris Johns: [kern/sleepqueue: Update due to API changes] http://devel.rtems.org/changeset/1bf6529/rtems-libbsd 2016-07-28T02:19:12 [git] a0ea5f8 by Chris Johns: [Revert "kern/sleepqueue: Update due to API changes" This reverts commit 1bf6529d195848be4f70b091a0a5071b1c08239a. It changes FreeBSD whitespace.] http://devel.rtems.org/changeset/a0ea5f8/rtems-libbsd 2016-07-28T02:19:12 [git] b6c170f by Chris Johns: [kern/sleepqueue: Update due to API changes] http://devel.rtems.org/changeset/b6c170f/rtems-libbsd 2016-07-28T03:40:44 *** jkridner has quit IRC 2016-07-28T03:41:19 *** jkridner has joined #rtems 2016-07-28T03:46:33 *** jkridner has quit IRC 2016-07-28T04:53:11 *** jkridner has joined #rtems 2016-07-28T04:58:53 *** jkridner has quit IRC 2016-07-28T05:49:51 *** monstr has joined #rtems 2016-07-28T06:07:51 [git] bf7faad by Sebastian Huber: [rtems-bsd-mutex: Update due to API changes] http://devel.rtems.org/changeset/bf7faad/rtems-libbsd 2016-07-28T06:48:22 sebhub: that cp15 works well 2016-07-28T06:48:29 cp15 call works well 2016-07-28T07:21:04 *** Wonjun has joined #rtems 2016-07-28T08:58:49 *** mumptai has quit IRC 2016-07-28T10:07:03 [git] b1ef367 by Sebastian Huber: [bsp/qoriq: Add QORIQ_HAS_WRITE_BACK_L1_CACHE Fixes start via U-Boot on P1020.] http://devel.rtems.org/changeset/b1ef367/rtems 2016-07-28T10:07:03 [git] 6446773 by Sebastian Huber: [bsp/qoriq: Fix tlbwe synchronization] http://devel.rtems.org/changeset/6446773/rtems 2016-07-28T10:07:03 [git] cecc109 by Sebastian Huber: [bsp/qoriq: Fix start stack setup] http://devel.rtems.org/changeset/cecc109/rtems 2016-07-28T10:07:03 [git] f383f4b by Sebastian Huber: [bsp/qoriq: Remove legacy U-Boot support] http://devel.rtems.org/changeset/f383f4b/rtems 2016-07-28T11:49:05 *** jkridner has joined #rtems 2016-07-28T11:50:16 *** mudit1729 has quit IRC 2016-07-28T12:16:18 *** BBD has joined #rtems 2016-07-28T12:16:32 Hello everybody 2016-07-28T12:22:54 I am currently working with RTEMS-libbsd ported on Beaglebone Black(BBB) and I might need the help of the community regarding an issue. I managed to port the cpsw driver from FreeBSD to RTEMS-libbsd to make the BBB network interface available with is interrupts. Hence, to test the network interface, I am trying to do some ping tests to monitor interrupts behavior. Here is my issue: 2016-07-28T12:29:55 I am trying to configure the DHCPCD program to do a static configuration of the network with the dhcpcd.conf file. Notwithstanding, it seems that whether I define statically the routes or not, the network is unreachable. Extra info: the network I use has a domain name. Does this domain name have an impact with the current implementation ? 2016-07-28T12:30:38 Thanks in advance 2016-07-28T13:21:54 works the dhcpcd01 test program? is the driver all right? 2016-07-28T13:22:32 *** BBD has quit IRC 2016-07-28T15:59:47 *** monstr has quit IRC 2016-07-28T17:07:08 *** Davidbrcz has joined #rtems 2016-07-28T17:12:34 *** Davidbrcz has quit IRC 2016-07-28T17:23:22 *** dr__house|away is now known as dr__house 2016-07-28T17:26:38 *** Davidbrcz has joined #rtems 2016-07-28T17:39:12 *** gedare has joined #rtems 2016-07-28T17:39:12 *** ChanServ sets mode: +o gedare 2016-07-28T17:41:23 *** Davidbrcz has quit IRC 2016-07-28T17:45:10 *** gedare has quit IRC 2016-07-28T17:54:34 *** gigetoo has quit IRC 2016-07-28T17:55:31 *** gigetoo has joined #rtems 2016-07-28T18:20:25 *** gedare has joined #rtems 2016-07-28T18:20:25 *** ChanServ sets mode: +o gedare 2016-07-28T18:24:59 *** Davidbrcz has joined #rtems 2016-07-28T18:26:43 *** gedare has quit IRC 2016-07-28T18:42:02 *** jkridner|work has joined #rtems 2016-07-28T18:45:54 *** jkridner has quit IRC 2016-07-28T18:51:02 *** jkridner has joined #rtems 2016-07-28T18:54:12 *** jkridner|work has quit IRC 2016-07-28T19:46:36 *** Davidbrcz has quit IRC 2016-07-28T20:18:39 *** gedare has joined #rtems 2016-07-28T20:18:39 *** ChanServ sets mode: +o gedare 2016-07-28T20:23:28 *** itazzzz is now known as ita 2016-07-28T20:33:04 *** gedare has quit IRC 2016-07-28T20:40:05 *** Davidbrcz has joined #rtems 2016-07-28T21:00:06 *** jkridner has quit IRC 2016-07-28T21:00:46 *** jkridner has joined #rtems 2016-07-28T21:05:19 *** jkridner has quit IRC 2016-07-28T22:01:34 *** Davidbrcz has quit IRC 2016-07-28T23:01:27 *** jkridner has joined #rtems 2016-07-29T00:31:51 *** gedare has joined #rtems 2016-07-29T00:31:51 *** ChanServ sets mode: +o gedare 2016-07-29T00:37:43 *** gedare has quit IRC 2016-07-29T00:57:31 *** dr__house is now known as dr__house|away 2016-07-29T01:48:16 *** jkridner has quit IRC 2016-07-29T03:49:51 *** Wonjun_ has joined #rtems 2016-07-29T03:52:48 *** Wonjun has quit IRC 2016-07-29T05:11:15 [trac] #2767/enhancement (v:4.11) created by pggauvin (ARM PREL31 and TARGET2 relocation support) http://devel.rtems.org/ticket/2767 2016-07-29T05:12:39 [trac] rtems-libdl-arm-prel31-target2.patch () attached to #2767 by pggauvin http://devel.rtems.org/ticket/2767 2016-07-29T05:13:28 [trac] libdl-prel31-target2-test.tar.gz () attached to #2767 by pggauvin http://devel.rtems.org/ticket/2767 2016-07-29T05:25:20 *** monstr has joined #rtems 2016-07-29T06:14:31 *** ita is now known as ita|zzz 2016-07-29T08:32:45 sebhub: hi 2016-07-29T08:33:49 sebhub: an observation, when --enable-profiling is configured some tests do not print there end message and there is profile prints. 2016-07-29T09:02:41 *** Davidbrcz has joined #rtems 2016-07-29T09:06:39 *** Wonjun__ has joined #rtems 2016-07-29T09:09:46 *** Wonjun_ has quit IRC 2016-07-29T09:11:21 *** Wonjun__ has quit IRC 2016-07-29T09:13:20 *** Wonjun__ has joined #rtems 2016-07-29T09:14:29 *** Davidbrcz has quit IRC 2016-07-29T09:14:35 *** Wonjun_ has joined #rtems 2016-07-29T09:17:52 *** Wonjun__ has quit IRC 2016-07-29T09:22:25 *** Wonjun__ has joined #rtems 2016-07-29T09:25:40 *** Wonjun_ has quit IRC 2016-07-29T09:28:55 *** Davidbrcz has joined #rtems 2016-07-29T13:14:28 *** Davidbrcz has quit IRC 2016-07-29T13:40:15 *** Davidbrcz has joined #rtems 2016-07-29T13:57:30 *** monstr has quit IRC 2016-07-29T14:03:14 *** Davidbrcz has quit IRC 2016-07-29T14:17:34 *** Davidbrcz has joined #rtems 2016-07-29T14:34:43 *** gedare has joined #rtems 2016-07-29T14:34:43 *** ChanServ sets mode: +o gedare 2016-07-29T14:36:33 *** benjamint has quit IRC 2016-07-29T15:20:30 *** jkridner has joined #rtems 2016-07-29T15:23:02 *** jkridner has quit IRC 2016-07-29T15:37:18 *** benjamint has joined #rtems 2016-07-29T16:38:06 *** ita|zzz is now known as ita 2016-07-29T16:42:11 *** Davidbrcz has quit IRC 2016-07-29T17:14:46 [trac] #2732/defect (new) updated by Gedare Bloom (In [changeset:"842005e432136953d670e39cedc2d665ea949e7b/rtems"]: {{{ ...) http://devel.rtems.org/ticket/2732 2016-07-29T17:14:46 [git] 842005e by Gedare Bloom: [posix: nanosleep: optimize away a time conversion updates #2732] http://devel.rtems.org/changeset/842005e/rtems 2016-07-29T17:29:44 *** Davidbrcz has joined #rtems 2016-07-29T17:39:21 *** gedare has quit IRC 2016-07-29T17:44:14 *** dr__house|away is now known as dr__house 2016-07-29T19:08:19 *** Davidbrcz has quit IRC 2016-07-29T19:20:22 *** gedare has joined #rtems 2016-07-29T19:20:22 *** ChanServ sets mode: +o gedare 2016-07-29T19:23:05 *** Davidbrcz has joined #rtems 2016-07-29T20:33:59 *** dr__house is now known as dr__house|away 2016-07-29T20:36:32 *** gedare has quit IRC 2016-07-29T20:37:52 *** gedare has joined #rtems 2016-07-29T20:37:52 *** ChanServ sets mode: +o gedare 2016-07-29T22:05:08 *** mric has quit IRC 2016-07-29T22:38:47 *** Davidbrcz has quit IRC 2016-07-30T00:23:20 kiwichris: ping 2016-07-30T00:34:24 *** gedare has quit IRC 2016-07-30T08:59:27 *** Davidbrcz has joined #rtems 2016-07-30T09:32:14 *** Davidbrcz has quit IRC 2016-07-30T09:46:31 *** Davidbrcz has joined #rtems 2016-07-30T15:47:52 *** Davidbrcz has quit IRC 2016-07-30T16:00:31 *** Davidbrcz has joined #rtems 2016-07-30T16:58:45 *** Davidbrcz has joined #rtems 2016-07-30T17:04:54 *** Davidbrcz has quit IRC 2016-07-30T20:04:57 *** ita is now known as ita|away 2016-07-30T23:21:03 *** Wonjun__ has quit IRC 2016-07-31T08:29:07 *** guerby has quit IRC 2016-07-31T08:34:06 *** guerby has joined #rtems 2016-07-31T09:34:30 [git] 3d3ad4d by Pavel Pisa: [arm/raspberrypi: VideoCore access corrections in cache operation and more error checking. The first, mistake in buffer size computation for cache flush and invalidate has been corrected. GCC __attribute__( ( aligned( 64 ) ) ) should work and works for local variables. Code ensures right stack alignment. But attribute has to be 2016-07-31T09:34:30 moved to type declaration to...] http://devel.rtems.org/changeset/3d3ad4d/rtems 2016-07-31T09:37:33 *** Davidbrcz has joined #rtems 2016-07-31T10:24:18 *** Davidbrcz has quit IRC 2016-07-31T10:40:13 *** Davidbrcz has joined #rtems 2016-07-31T11:10:50 [git] 912a1a9 by Pavel Pisa: [arm/raspberrypi: framebuffer driver checks for initialization and tries runtime init if required. Code works but there is problem how to setup caching for the regions (arm_cp15_set_translation_table_entries arguments). If ARMV7_MMU_DATA_READ_WRITE_CACHED is used then CPU accesses are noncoherent with VideoCore and some part of 2016-07-31T11:10:51 image update are not visible...] http://devel.rtems.org/changeset/912a1a9/rtems 2016-07-31T11:48:50 [trac] GSoC/2016 edited by Mudit (empty comment) http://devel.rtems.org/wiki/GSoC/2016 2016-07-31T12:51:17 *** Davidbrcz has quit IRC 2016-07-31T16:54:07 *** ita|away is now known as ita 2016-07-31T19:34:53 *** mumptai has joined #rtems 2016-07-31T23:05:40 *** gedare has joined #rtems 2016-07-31T23:05:40 *** ChanServ sets mode: +o gedare 2016-07-31T23:30:30 *** gedare has quit IRC