linaro-dev

linaro-dev@lists.linaro.org

4332 discussions

by Harigopal Gollamudi

== Progress == DEVPLAT-216: Upgrading the ARM trusted firmware with fip binaries. Able to boot the kernel. Got stuck at not mounting the root file system to boot android. == Plans == DEVPLAT-216: Get boot android on fastmodel with FIP binaries == Issues == None

11 years, 3 months

[ACTIVITY] week 21

by Dileep Katta

== Progress == * Enabling UEFI Fastboot on Panda (CARD-198: 60 %) * Enabling U-boot Fastboot for Panda (CARD-1103: 20%) * Flash support on u-boot fastboot - Analysis (CARD-1103: 10%) * Familiarization of ubuntu fastboot tools (10%) == Plans == * Enabling UEFI Fastboot on Panda (CARD-198: 40 %) * Enabling U-boot Fastboot for Panda (CARD-1103: 20%) * Flash support on u-boot fastboot (CARD-1103: 40%) == Issues == * None

11 years, 3 months

[ACTIVITY] week 21

by Bernhard Rosenkränzer

== Progress == * Integrated last bits of new ARMv8 string handling into Bionic (JUICE-298: 10%) * Upstreamed clang 3.5 related fixes to AOSP (JUICE-340: 20%) * Followed up on gcc 4.9 bugs w/ Google (JUICE-341: 5%) * Looked into cbuildv2 for toolchain release builds (CARD-790: 30%) * Investigated libutils/stlport related crashes probably caused by gcc 4.9 (no card yet, 30%) * Helped Yongqin look at busybox build failure on Android armv8 (no card. 5%) == Plans == * Toolchain 2014.05 binary release (CARD-790) * Port new Bionic optimizations already done for ARMv8 to Cortex-A15 (DEVPLAT-104) * Fix libutils/stlport related crashes (no card yet) == Issues == * None

11 years, 3 months

[ACTIVITY] week 21

by Vishal Bhoj

== Progress == * Boot Android with u-boot on ARMv8 SoC (CARD-1049 80%) * CTS execution on Vexpress-LSK (REQ-280 10%) * JAVA 7 support in android-build (bug1310939 10%) == Plans == * Boot Android with u-boot on ARMv8 SoC. * 14.05 Android Release. == Issues == * None

11 years, 3 months

[ACTIVITY] week 21

by Riku Voipio

== Progress == * Xen-base ARMv8 (CARD-736 10%) * Xen packaging fix (BUG-1317155 20%) * Nova CI bringup (CARD-1211 10%) * Other CI fixes/workarounds 40% * Started DTB install location discussion at cross-distro == Plans == * Concentrate in Xen ARMv8 CI (CARD-736) * KVM ARMv8 CI (CARD-346) == Issues == * none

11 years, 3 months

[ACTIVITY] week 21

by Koen Kooi

== Progress == * Fix OE merge window build breakages. (CARD-787: 50%) * Moved DISTRO type items into meta-linaro-integration. (CARD-1214: 40%) * On-target git fixes (5%) * Release scripts (5%) == Plans == * Close out CARD-1214 * Release Linaro 14.05 == Issues == * None -- Koen Kooi Builds and Baselines | Release Manager Linaro.org | Open source software for ARM SoCs

11 years, 3 months

Adding UEFI support to BeagleBone Black | Introducing Myself

by Varad Gautam

Hi everyone, I am working to port UEFI (EDKII) to the BeagleBone Black as a Google Summer of Code intern with Leif Lindolm of Linaro Enterprise Group. Over the past couple of weeks I've been going through the EDKII Module Writer's Guide, the build file specifications, the SoC and board TRMs and the edk2 repo bundled documentation and got my BBB running with ssh / vnc / ftdi debugger set up. I have broken down the implementation into two parts, based on [1]: * Create an Am335xPkg to support the ARM SoC, taking ArmPlatformPkg as base and changing the build spec files and library implementations. * Use this to build BeagleBoneBlackPkg. = Questions = I am a little unclear about which Libraries and Drivers from the ArmPlatformPkg and ArmPkg I will require in the port, and which of these I can use as-provided from the repo. How do I figure this out? Also, is there any more documentation I should look into related to UEFI porting? I will be documenting my work at [2] (nothing relevant there yet) and will soon set up a github repo. I'll keep you posted about my progress. Thanks, Varad [1] http://sourceforge.net/apps/mediawiki/tianocore/index.php?title=ArmPlatform… [2] varadgautam.wordpress.com

11 years, 3 months

[PATCH] pm-qa/powertop test fails on ARM

by Mohammad Merajul Islam Molla

Hello, pm-qa/powertop test fails on ARM, even though it succeeds for Intel. on both platforms, powertop installed using command - "sudo apt-get install powertop" powertop_01.1: checking if reports are generated... Ok [on Intel] powertop_01.1: checking if reports are generated... Err [on ARM] It looks like "report_name=PowerTOP*.csv" does not work for ARM. On ARM, report name are lowercase. So report_name should be "powertop*.csv". If I build from git from (git.linaro.org), generated report file name is powertop.csv. So even then it needs to be lowercase. I made a patch to make it work on both platforms (even though the actual fix is probably to make the output filenames similar in both cases). Please comment. --------- diff --git a/powertop/powertop_01.sh b/powertop/powertop_01.sh index c66b725..2768924 100755 --- a/powertop/powertop_01.sh +++ b/powertop/powertop_01.sh @@ -34,6 +34,12 @@ run_powertop() { local seconds=10 local iterations=2 local report_name=PowerTOP*.csv + local processor=`uname -p` + + case "$processor" in + arm*) + report_name=powertop*.csv + esac # remove old reports if exists rm -f $report_name -- Thanks, -Meraj

11 years, 3 months

pm-qa/heat_cpu.c: Some Issues

by Mohammad Merajul Islam Molla

Hello Amit/Robert/Others, I was looking into the code for heat_cpu.c in pm-qa. I think there are some issues with the current implementation. Please find details below - 1. #define NR_THREAD 3 pid_t thr_id[NR_THREAD]; I think the array thr_id should be dynamically allocated as number of threads depend on number of cpus. However, we can eliminate the need for this array as mentioned below at no 7. 2. There is no error checking for sysconf (...) call (line 83). If this call fails num_cpus will be -1 which is a problem. 3. at line 122, pthread_t *p_thread_ptr = (pthread_t *) malloc( num_cpus * sizeof(pthread_attr_t)); we are allocating space for pthread_t. So sizeof(ptread_t) should be used instead of sizeof(pthread_attr_t). 4. at line 144, SCHED_NORMAL is used. On my Ubuntu 13.04, <sched.h> does not define SCHED_NORMAL. So it gives compile error if ANDROID is defined. heat_cpu.c:145:44: error: ‘SCHED_NORMAL’ undeclared (first use in this function) We can either use SCHED_OTHER (which is equivalent to SCHED_NORMAL) or include <linux/sched.h> 5. at line 168, error message is misleading. This error is related to failure from pthread_create, not during setting affinity. 6. at line 76, the call to gettid results in linking error - heat_cpu.c: In function ‘do_loop’: heat_cpu.c:77:2: warning: implicit declaration of function ‘gettid’ [-Wimplicit-function-declaration] /tmp/ccAhcUGE.o: In function `do_loop': heat_cpu.c:(.text+0x4e): undefined reference to `gettid' There is no libc wrapper for gettid. So I think it should be called as below - #include <sys/syscall.h> thr_id[thread] = syscall(SYS_gettid); With this changes the compilation warning and link error goes away and it works. However, this call can be eliminated as suggested below at no 7. 7. There is a potential race condition between line 171 & 182, despite the synchronization mechanism (mutex & condition variable) used. There is no guaranty that do_loop(...) will execute first and thr_id[thread] = gettid(); will execute before setting affinity as noted from below debug output - setting affinity for : 0 setting affinity for : 0 setting affinity for : 0 setting affinity for : 0 The code is setting affinity for tid 0, which it should not, despite the synchronization code. However, we can eliminate need for thr_id and the associated synchronization if we use pthread_setaffinity_np(...) . Would you care to test the alternative implementation using the patch below (also attached)? -------------------------------------------------------------------- diff --git a/utils/heat_cpu.c b/utils/heat_cpu.c index 90d8a97..236cb1d 100644 --- a/utils/heat_cpu.c +++ b/utils/heat_cpu.c @@ -47,23 +47,12 @@ #include <string.h> #include <math.h> -#define NR_THREAD 3 - int cont = 1; -int cpu_index; long long a = 1, c = 1; float f = M_PI; -pid_t thr_id[NR_THREAD]; int moderate_inst; - -#ifdef ANDROID -int conditionMet; -pthread_cond_t cond = PTHREAD_COND_INITIALIZER; -pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER; -#endif - void *do_loop(void *data) { const long long b = 3; @@ -71,14 +60,6 @@ void *do_loop(void *data) a = 1; c = 1; -#ifdef ANDROID - long int thread = (long int) data; - thr_id[thread] = gettid(); - pthread_mutex_lock(&mutex); - while (!conditionMet) - pthread_cond_wait(&cond, &mutex); - pthread_mutex_unlock(&mutex); -#endif while (cont) { if (!moderate_inst) { @@ -96,9 +77,15 @@ void *do_loop(void *data) int main(int arg_count, char *argv[]) { int ret, i; - int num_cpus = sysconf(_SC_NPROCESSORS_ONLN); + int num_cpus; cpu_set_t cpuset; + num_cpus = sysconf(_SC_NPROCESSORS_ONLN); + if (num_cpus < 0) { + printf("ERROR: sysconf error\n"); + return -1; + } + printf("Num CPUs: %d\n", num_cpus); if (arg_count > 1) { if (!strcmp("moderate", argv[1])) { @@ -120,7 +107,7 @@ int main(int arg_count, char *argv[]) } pthread_t *p_thread_ptr = (pthread_t *) malloc( - num_cpus * sizeof(pthread_attr_t)); + num_cpus * sizeof(pthread_t)); if (!p_thread_ptr) { printf("ERROR: out of memory\n"); @@ -141,7 +128,7 @@ int main(int arg_count, char *argv[]) #endif #else - ret = pthread_attr_setschedpolicy(&p[i], SCHED_NORMAL); + ret = pthread_attr_setschedpolicy(&p[i], SCHED_OTHER); #endif /* for each new object */ CPU_SET(i, &cpuset); @@ -156,7 +143,7 @@ int main(int arg_count, char *argv[]) return ret; } #endif - CPU_CLR(i, &cpuset); + CPU_CLR(i, &cpuset); } @@ -165,29 +152,23 @@ int main(int arg_count, char *argv[]) ret = pthread_create(&p_thread_ptr[i], &p[i], do_loop, (void *)i); if (ret < 0) - printf("Error setting affinity for cpu%d\n", i); + printf("Error creating thread for cpu%d\n", i); #ifdef ANDROID CPU_ZERO(&cpuset); CPU_SET(i, &cpuset); - ret = sched_setaffinity(thr_id[i], sizeof(cpuset), &cpuset); + ret = pthread_setaffinity_np(p_thread_ptr[i], sizeof(cpuset), &cpuset); if (ret) { - printf("Error setting affinity on pthread th_id\n"); + printf("Error setting affinity on pthread\n"); printf("Error: %s\n", strerror(ret)); return ret; } #endif } -#ifdef ANDROID - pthread_mutex_lock(&mutex); - conditionMet = 1; - pthread_cond_broadcast(&cond); - pthread_mutex_unlock(&mutex); -#endif - while (1) sleep(1); + return 0; } -- Thanks, -Meraj

11 years, 4 months

[PATCH PMQA] cpuidle_killer: allocate per cpu process array dynamically

by Sanjay Singh Rawat

currently percpu process array is set to 2, which results in segfault Signed-off-by: Sanjay Singh Rawat <sanjay.rawat(a)linaro.org> --- cpuidle/cpuidle_killer.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/cpuidle/cpuidle_killer.c b/cpuidle/cpuidle_killer.c index 5e7320f..09009ef 100644 --- a/cpuidle/cpuidle_killer.c +++ b/cpuidle/cpuidle_killer.c @@ -104,7 +104,7 @@ int main(int argc, char *argv[]) { int ret, i, nrcpus = 2; int nrsleeps, delay; - pid_t pids[nrcpus]; + pid_t *pids; struct timex timex = { 0 }; if (adjtimex(&timex) < 0) { @@ -121,6 +121,11 @@ int main(int argc, char *argv[]) } fprintf(stderr, "found %d cpu(s)\n", nrcpus); + pids = (pid_t *) calloc(nrcpus, sizeof(pid_t)); + if (pids == NULL) { + fprintf(stderr, "error: calloc failed\n"); + return 1; + } for (i = 0; i < nrcpus; i++) { -- 1.7.10.4

11 years, 4 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

linaro-dev