Hello,
I'm running https://android-build.linaro.org/builds/~linaro-android/panda-jb-gcc47-tilt-..., build 23 (Aug15). It's kernel version is 3.4.0-xxx, but 'perf --version' reports 3.0.8_android. Why such a version mismatch? Usually it's the very poor idea to run perf on top of mismatched kernel. For the particular build, any attempt to run something beyond 'perf --help' or 'perf --version' crashes everything, for example:
[ 65.116027] Unable to handle kernel paging request at virtual address e66107c0 [ 65.123687] pgd = e4664000 [ 65.126525] [e66107c0] *pgd=00000000 [ 65.130310] Internal error: Oops: 5 [#1] PREEMPT SMP ARM [ 65.135955] Modules linked in: wl12xx wlcore mac80211 cfg80211 lib80211 btwilink rfcomm bnep bluetooth gator [ 65.146423] CPU: 0 Not tainted (3.4.0-gcc0d09f #1) [ 65.151885] PC is at vmap_page_range_noflush+0xc8/0x19c [ 65.157409] LR is at map_kernel_range_noflush+0x1c/0x20 [ 65.162963] pc : [<c00e5b28>] lr : [<c00e7018>] psr: 80030113 [ 65.162963] sp : e3ff5d80 ip : e66107c0 fp : e3ff5dc4 [ 65.175109] r10: e66107c0 r9 : c0725048 r8 : c07ba0c0 [ 65.180633] r7 : c0007f98 r6 : fe7f1000 r5 : 00000000 r4 : fe7f0000 [ 65.187622] r3 : e3900c40 r2 : e66107c0 r1 : fe7f0fff r0 : fe7f0000 [ 65.194549] Flags: Nzcv IRQs on FIQs on Mode SVC_32 ISA ARM Segment user [ 65.202087] Control: 10c5387d Table: a466404a DAC: 00000015 <...registers...> [ 66.326110] Process perf (pid: 1811, stack limit = 0xe3ff42f8) [ 66.332244] Stack: (0xe3ff5d80 to 0xe3ff6000) [ 66.336853] 5d80: e3ff5e0c 00000000 c0779c44 e3900c40 0000065f fe7f1000 c00d69fc e3900f40 [ 66.345489] 5da0: 00000800 00000000 00000000 00000000 e3900940 00000000 e3ff5dd4 e3ff5dc8 [ 66.354125] 5dc0: c00e7018 c00e5a6c e3ff5e44 e3ff5dd8 c00d7b5c c00e7008 e3ff5e34 e3ff5de8 [ 66.362762] 5de0: fe7f0000 00001000 e4666e20 c0721ea8 00000001 00000001 00000004 e3900c40 [ 66.371398] 5e00: 00000001 00000000 e3fb2800 e3900940 00000000 00000001 00000001 00000004 [ 66.380035] 5e20: c07b5510 c0721efc 00000000 e3fb2800 c07b5670 c074e510 e3ff5e54 e3ff5e48 [ 66.388641] 5e40: c00d7fa0 c00d76a8 e3ff5e7c e3ff5e58 c00ab740 c00d7f90 00000800 00000000 [ 66.397277] 5e60: e3fb2800 00000000 e3fb2800 e3ff5f28 e3ff5e9c e3ff5e80 c00b0fe8 c00ab5b8 [ 66.405914] 5e80: c00b0fac c0750924 e3fb2800 c0750814 e3ff5ec4 e3ff5ea0 c00b7854 c00b0fb8 [ 66.414550] 5ea0: 00000000 c0750924 00000001 00000000 00000000 00000000 e3ff5ef4 e3ff5ec8 [ 66.423187] 5ec0: c00b7b04 c00b77e8 00000000 00000000 00000000 ffffffff 00000000 00000000 [ 66.431823] 5ee0: 00000000 00000000 e3ff5fa4 e3ff5ef8 c00b8458 c00b78b0 00000000 00000000 [ 66.440460] 5f00: 00000000 c001a188 e40a7590 00000010 e3ff5f2c e3ff5f20 00000003 00000000 [ 66.449096] 5f20: e3ff5f64 00000000 00000002 00000048 00000037 00000000 00000001 00000000 [ 66.457733] 5f40: 000004c7 00000000 00000007 00000000 00040302 00000000 00000000 00000000 [ 66.466339] 5f60: 00000000 00000000 00000000 00000000 00000000 00000000 e3ff5fa4 00000000 [ 66.474975] 5f80: 00000000 ffffffff 0000016c c000e6e8 e3ff4000 00000000 00000000 e3ff5fa8 [ 66.483612] 5fa0: c000e480 c00b8024 00000000 00000000 b899c630 ffffffff 00000000 ffffffff [ 66.492248] 5fc0: 00000000 00000000 ffffffff 0000016c b899c628 b899d160 00000008 b899e580 [ 66.500885] 5fe0: be8f3870 be8f3860 b6f886a9 b6eecef0 20030110 b899c630 00000000 20000000 [ 66.509521] Backtrace: [ 66.512115] [<c00e5a60>] (vmap_page_range_noflush+0x0/0x19c) from [<c00e7018>] (map_kernel_range_noflush+0x1c/0x20) [ 66.523162] [<c00e6ffc>] (map_kernel_range_noflush+0x0/0x20) from [<c00d7b5c>] (pcpu_alloc+0x4c0/0x8e8) [ 66.533081] [<c00d769c>] (pcpu_alloc+0x0/0x8e8) from [<c00d7fa0>] (__alloc_percpu+0x1c/0x20) [ 66.541992] [<c00d7f84>] (__alloc_percpu+0x0/0x20) from [<c00ab740>] (perf_trace_init+0x194/0x2c4) [ 66.551483] [<c00ab5ac>] (perf_trace_init+0x0/0x2c4) from [<c00b0fe8>] (perf_tp_event_init+0x3c/0x60) [ 66.561218] [<c00b0fac>] (perf_tp_event_init+0x0/0x60) from [<c00b7854>] (perf_init_event+0x78/0xc8) [ 66.570861] r6:c0750814 r5:e3fb2800 r4:c0750924 r3:c00b0fac [ 66.576843] [<c00b77dc>] (perf_init_event+0x0/0xc8) from [<c00b7b04>] (perf_event_alloc+0x260/0x420) [ 66.586486] r7:00000000 r6:00000000 r5:00000000 r4:00000001 [ 66.592498] [<c00b78a4>] (perf_event_alloc+0x0/0x420) from [<c00b8458>] (sys_perf_event_open+0x440/0x818) [ 66.602600] [<c00b8018>] (sys_perf_event_open+0x0/0x818) from [<c000e480>] (ret_fast_syscall+0x0/0x30) [ 66.612426] Code: e59f90d4 e1a0a00c e1a0200a e51b3038 (e592e000) [ 66.619354] ---[ end trace 8f2771e9f1c2026e ]--- [ 66.624450] Kernel panic - not syncing: Fatal exception [ 66.630035] CPU1: stopping [ 66.632904] Backtrace: [ 66.635498] [<c0012a04>] (dump_backtrace+0x0/0x114) from [<c04e0f90>] (dump_stack+0x20/0x24) [ 66.644409] r6:00000000 r5:fa240100 r4:00000001 r3:c074bccc [ 66.650421] [<c04e0f70>] (dump_stack+0x0/0x24) from [<c00141ec>] (handle_IPI+0x10c/0x1e0) [ 66.659057] [<c00140e0>] (handle_IPI+0x0/0x1e0) from [<c0008678>] (gic_handle_irq+0x68/0x70) [ 66.667968] r7:e609ff9c r6:c0721900 r5:fa240100 r4:e609ff68 [ 66.673980] [<c0008610>] (gic_handle_irq+0x0/0x70) from [<c000e080>] (__irq_svc+0x40/0x70) [ 66.682678] Exception stack(0xe609ff68 to 0xe609ffb0) [ 66.688018] ff60: c077c428 27aa89aa 4030a263 fe404000 e609e000 c04ee620 [ 66.696655] ff80: c077c270 c077c284 8000406a 411fc092 00000000 e609ffbc e609ff9c e609ffb0 [ 66.705291] ffa0: c0025384 c000f7a8 a0010113 ffffffff [ 66.710632] r6:ffffffff r5:a0010113 r4:c000f7a8 r3:c0025384 [ 66.716613] [<c000f774>] (default_idle+0x0/0x3c) from [<c000fb94>] (cpu_idle+0xbc/0x114) [ 66.725158] [<c000fad8>] (cpu_idle+0x0/0x114) from [<c04ddd94>] (secondary_start_kernel+0x130/0x154) [ 66.734802] r5:00000001 r4:00000001 [ 66.738616] [<c04ddc64>] (secondary_start_kernel+0x0/0x154) from [<804dd6d4>] (0x804dd6d4) [ 66.747314] r6:10c0387d r5:00000015 r4:a608c06a r3:c04dd6bc [ 66.753326] panic occurred, switching back to text console
Dmitry
On 08/15/12 16:23, the mail apparently from Dmitry Antipov included:
Hello,
I'm running https://android-build.linaro.org/builds/~linaro-android/panda-jb-gcc47-tilt-...,
build 23 (Aug15). It's kernel version is 3.4.0-xxx, but 'perf --version' reports 3.0.8_android. Why such a version mismatch? Usually it's the very poor idea to run perf on top of mismatched
Indeed.
The Android guys package it and provide perf, we don't have anything to do with that side of it. Vishal probably has a better idea.
-Andy
kernel. For the particular build, any attempt to run something beyond 'perf --help' or 'perf --version' crashes everything, for example:
[ 65.116027] Unable to handle kernel paging request at virtual address e66107c0 [ 65.123687] pgd = e4664000 [ 65.126525] [e66107c0] *pgd=00000000 [ 65.130310] Internal error: Oops: 5 [#1] PREEMPT SMP ARM [ 65.135955] Modules linked in: wl12xx wlcore mac80211 cfg80211 lib80211 btwilink rfcomm bnep bluetooth gator [ 65.146423] CPU: 0 Not tainted (3.4.0-gcc0d09f #1) [ 65.151885] PC is at vmap_page_range_noflush+0xc8/0x19c [ 65.157409] LR is at map_kernel_range_noflush+0x1c/0x20 [ 65.162963] pc : [<c00e5b28>] lr : [<c00e7018>] psr: 80030113 [ 65.162963] sp : e3ff5d80 ip : e66107c0 fp : e3ff5dc4 [ 65.175109] r10: e66107c0 r9 : c0725048 r8 : c07ba0c0 [ 65.180633] r7 : c0007f98 r6 : fe7f1000 r5 : 00000000 r4 : fe7f0000 [ 65.187622] r3 : e3900c40 r2 : e66107c0 r1 : fe7f0fff r0 : fe7f0000 [ 65.194549] Flags: Nzcv IRQs on FIQs on Mode SVC_32 ISA ARM Segment user [ 65.202087] Control: 10c5387d Table: a466404a DAC: 00000015 <...registers...> [ 66.326110] Process perf (pid: 1811, stack limit = 0xe3ff42f8) [ 66.332244] Stack: (0xe3ff5d80 to 0xe3ff6000) [ 66.336853] 5d80: e3ff5e0c 00000000 c0779c44 e3900c40 0000065f fe7f1000 c00d69fc e3900f40 [ 66.345489] 5da0: 00000800 00000000 00000000 00000000 e3900940 00000000 e3ff5dd4 e3ff5dc8 [ 66.354125] 5dc0: c00e7018 c00e5a6c e3ff5e44 e3ff5dd8 c00d7b5c c00e7008 e3ff5e34 e3ff5de8 [ 66.362762] 5de0: fe7f0000 00001000 e4666e20 c0721ea8 00000001 00000001 00000004 e3900c40 [ 66.371398] 5e00: 00000001 00000000 e3fb2800 e3900940 00000000 00000001 00000001 00000004 [ 66.380035] 5e20: c07b5510 c0721efc 00000000 e3fb2800 c07b5670 c074e510 e3ff5e54 e3ff5e48 [ 66.388641] 5e40: c00d7fa0 c00d76a8 e3ff5e7c e3ff5e58 c00ab740 c00d7f90 00000800 00000000 [ 66.397277] 5e60: e3fb2800 00000000 e3fb2800 e3ff5f28 e3ff5e9c e3ff5e80 c00b0fe8 c00ab5b8 [ 66.405914] 5e80: c00b0fac c0750924 e3fb2800 c0750814 e3ff5ec4 e3ff5ea0 c00b7854 c00b0fb8 [ 66.414550] 5ea0: 00000000 c0750924 00000001 00000000 00000000 00000000 e3ff5ef4 e3ff5ec8 [ 66.423187] 5ec0: c00b7b04 c00b77e8 00000000 00000000 00000000 ffffffff 00000000 00000000 [ 66.431823] 5ee0: 00000000 00000000 e3ff5fa4 e3ff5ef8 c00b8458 c00b78b0 00000000 00000000 [ 66.440460] 5f00: 00000000 c001a188 e40a7590 00000010 e3ff5f2c e3ff5f20 00000003 00000000 [ 66.449096] 5f20: e3ff5f64 00000000 00000002 00000048 00000037 00000000 00000001 00000000 [ 66.457733] 5f40: 000004c7 00000000 00000007 00000000 00040302 00000000 00000000 00000000 [ 66.466339] 5f60: 00000000 00000000 00000000 00000000 00000000 00000000 e3ff5fa4 00000000 [ 66.474975] 5f80: 00000000 ffffffff 0000016c c000e6e8 e3ff4000 00000000 00000000 e3ff5fa8 [ 66.483612] 5fa0: c000e480 c00b8024 00000000 00000000 b899c630 ffffffff 00000000 ffffffff [ 66.492248] 5fc0: 00000000 00000000 ffffffff 0000016c b899c628 b899d160 00000008 b899e580 [ 66.500885] 5fe0: be8f3870 be8f3860 b6f886a9 b6eecef0 20030110 b899c630 00000000 20000000 [ 66.509521] Backtrace: [ 66.512115] [<c00e5a60>] (vmap_page_range_noflush+0x0/0x19c) from [<c00e7018>] (map_kernel_range_noflush+0x1c/0x20) [ 66.523162] [<c00e6ffc>] (map_kernel_range_noflush+0x0/0x20) from [<c00d7b5c>] (pcpu_alloc+0x4c0/0x8e8) [ 66.533081] [<c00d769c>] (pcpu_alloc+0x0/0x8e8) from [<c00d7fa0>] (__alloc_percpu+0x1c/0x20) [ 66.541992] [<c00d7f84>] (__alloc_percpu+0x0/0x20) from [<c00ab740>] (perf_trace_init+0x194/0x2c4) [ 66.551483] [<c00ab5ac>] (perf_trace_init+0x0/0x2c4) from [<c00b0fe8>] (perf_tp_event_init+0x3c/0x60) [ 66.561218] [<c00b0fac>] (perf_tp_event_init+0x0/0x60) from [<c00b7854>] (perf_init_event+0x78/0xc8) [ 66.570861] r6:c0750814 r5:e3fb2800 r4:c0750924 r3:c00b0fac [ 66.576843] [<c00b77dc>] (perf_init_event+0x0/0xc8) from [<c00b7b04>] (perf_event_alloc+0x260/0x420) [ 66.586486] r7:00000000 r6:00000000 r5:00000000 r4:00000001 [ 66.592498] [<c00b78a4>] (perf_event_alloc+0x0/0x420) from [<c00b8458>] (sys_perf_event_open+0x440/0x818) [ 66.602600] [<c00b8018>] (sys_perf_event_open+0x0/0x818) from [<c000e480>] (ret_fast_syscall+0x0/0x30) [ 66.612426] Code: e59f90d4 e1a0a00c e1a0200a e51b3038 (e592e000) [ 66.619354] ---[ end trace 8f2771e9f1c2026e ]--- [ 66.624450] Kernel panic - not syncing: Fatal exception [ 66.630035] CPU1: stopping [ 66.632904] Backtrace: [ 66.635498] [<c0012a04>] (dump_backtrace+0x0/0x114) from [<c04e0f90>] (dump_stack+0x20/0x24) [ 66.644409] r6:00000000 r5:fa240100 r4:00000001 r3:c074bccc [ 66.650421] [<c04e0f70>] (dump_stack+0x0/0x24) from [<c00141ec>] (handle_IPI+0x10c/0x1e0) [ 66.659057] [<c00140e0>] (handle_IPI+0x0/0x1e0) from [<c0008678>] (gic_handle_irq+0x68/0x70) [ 66.667968] r7:e609ff9c r6:c0721900 r5:fa240100 r4:e609ff68 [ 66.673980] [<c0008610>] (gic_handle_irq+0x0/0x70) from [<c000e080>] (__irq_svc+0x40/0x70) [ 66.682678] Exception stack(0xe609ff68 to 0xe609ffb0) [ 66.688018] ff60: c077c428 27aa89aa 4030a263 fe404000 e609e000 c04ee620 [ 66.696655] ff80: c077c270 c077c284 8000406a 411fc092 00000000 e609ffbc e609ff9c e609ffb0 [ 66.705291] ffa0: c0025384 c000f7a8 a0010113 ffffffff [ 66.710632] r6:ffffffff r5:a0010113 r4:c000f7a8 r3:c0025384 [ 66.716613] [<c000f774>] (default_idle+0x0/0x3c) from [<c000fb94>] (cpu_idle+0xbc/0x114) [ 66.725158] [<c000fad8>] (cpu_idle+0x0/0x114) from [<c04ddd94>] (secondary_start_kernel+0x130/0x154) [ 66.734802] r5:00000001 r4:00000001 [ 66.738616] [<c04ddc64>] (secondary_start_kernel+0x0/0x154) from [<804dd6d4>] (0x804dd6d4) [ 66.747314] r6:10c0387d r5:00000015 r4:a608c06a r3:c04dd6bc [ 66.753326] panic occurred, switching back to text console
Dmitry
Hello Vishal,
is it possible to package the perf version matched to the kernel version for our Android builds? Otherwise perf is just useless.
Dmitry
Hi,
On 15 August 2012 10:23, Dmitry Antipov dmitry.antipov@linaro.org wrote:
I'm running https://android-build.linaro.org/builds/~linaro-android/panda-jb-gcc47-tilt-..., build 23 (Aug15). It's kernel version is 3.4.0-xxx, but 'perf --version' reports 3.0.8_android. Why such a version mismatch?
The upstream Android guys decided to take the easy way and fork an old version of perf instead of getting Android support into the version included in the kernel, so in normal JB userland, you'll always have perf 3.0.8 even if you're running kernel 5.0.
I've fixed it up, we have a version of current perf that works w/ Android - but it's not yet in the builds. (I've fixed it the "proper" way by patching the kernel tree and submitting the patch - but currently Android's external perf is the one that gets installed)
ttyl bero
On Wed, Aug 15, 2012 at 4:38 PM, Bernhard Rosenkränzer bernhard.rosenkranzer@linaro.org wrote:
Hi,
On 15 August 2012 10:23, Dmitry Antipov dmitry.antipov@linaro.org wrote:
I'm running https://android-build.linaro.org/builds/~linaro-android/panda-jb-gcc47-tilt-..., build 23 (Aug15). It's kernel version is 3.4.0-xxx, but 'perf --version' reports 3.0.8_android. Why such a version mismatch?
The upstream Android guys decided to take the easy way and fork an old version of perf instead of getting Android support into the version included in the kernel, so in normal JB userland, you'll always have perf 3.0.8 even if you're running kernel 5.0.
I've fixed it up, we have a version of current perf that works w/ Android - but it's not yet in the builds. (I've fixed it the "proper" way by patching the kernel tree and submitting the patch - but currently Android's external perf is the one that gets installed)
Thanks for this Bero. How soon before this gets into a build we can use?
/Amit
Hi,
On 15 August 2012 13:25, Amit Kucheria amit.kucheria@linaro.org wrote:
Thanks for this Bero. How soon before this gets into a build we can use?
We're planning on pulling it in through upstream, so (unless it's a priority and we start pushing it in ourselves), it's a matter of waiting for the patch to be accepted.
ttyl bero
On Wed, Aug 15, 2012 at 7:07 PM, Bernhard Rosenkränzer bernhard.rosenkranzer@linaro.org wrote:
Hi,
On 15 August 2012 13:25, Amit Kucheria amit.kucheria@linaro.org wrote:
Thanks for this Bero. How soon before this gets into a build we can use?
We're planning on pulling it in through upstream, so (unless it's a priority and we start pushing it in ourselves), it's a matter of waiting for the patch to be accepted.
Working perf is a priority else we'll be forced to revert back to using Ubuntu LEBs for creating mobile workloads and using those for our testing and analysis.
On 15 August 2012 09:23, Dmitry Antipov dmitry.antipov@linaro.org wrote:
Hello,
I'm running https://android-build.linaro.org/builds/~linaro-android/panda-jb-gcc47-tilt-..., build 23 (Aug15). It's kernel version is 3.4.0-xxx, but 'perf --version' reports 3.0.8_android. Why such a version mismatch? Usually it's the very poor idea to run perf on top of mismatched
Is that really true. Isn't it supposed to be backwards compatible ?
Ramana
On 08/15/2012 03:15 PM, Ramana Radhakrishnan wrote:
On 15 August 2012 09:23, Dmitry Antipov dmitry.antipov@linaro.org wrote:
Hello,
I'm running https://android-build.linaro.org/builds/~linaro-android/panda-jb-gcc47-tilt-..., build 23 (Aug15). It's kernel version is 3.4.0-xxx, but 'perf --version' reports 3.0.8_android. Why such a version mismatch? Usually it's the very poor idea to run perf on top of mismatched
Is that really true. Isn't it supposed to be backwards compatible ?
The perf.data format is designed with some compatibility in mind, so 3.0.8_android should read any recent enough (say, from perf >= 3.0) perf.data at least. But if perf can't record anything due to a wrong kernel version, it's almost useless.
Dmitry
-
Amit Kucheria -
Andy Green -
Bernhard Rosenkränzer -
Dmitry Antipov -
Ramana Radhakrishnan