Coresight ETF trace dump failed in Juno r1 with 4.8-rc8
Suzuki K Poulose
Suzuki.Poulose at arm.com
Wed Oct 5 10:17:05 PDT 2016
On 05/10/16 06:27, Venkatesh Vivekanandan wrote:
> On Tue, Oct 4, 2016 at 7:59 PM, Suzuki K Poulose <Suzuki.Poulose at arm.com> wrote:
>> On 04/10/16 06:37, Venkatesh Vivekanandan wrote:
>>>
>>> On Mon, Oct 3, 2016 at 6:44 PM, Sudeep Holla <sudeep.holla at arm.com> wrote:
>>>>
>>>> Hi Venkatesh,
>>>>
>>>> On 03/10/16 12:36, Venkatesh Vivekanandan wrote:
>>>>>
>>>>>
>>>>> Hi All,
>>>>>
>>>>> I am trying to collect ETF trace from Juno R1 and could see "cpu
>>>>> stall" while dumping the trace. Attached is the log of sequence
>>>>> followed. Was trying to collect the trace data from hardware and see
>>>>> if it is any valid data. Am I missing anything here?.
>>>>>
>>>>
>>>> There are few fixes from me and Suzuki queued for v4.9.
>>>> Can you check if this issue persists even on linux-next ?
>>>
>>>
>>> Issue is the same in linux-next as well. Please find the attached log.
>>>
>>> linaro-test [rc=0]# dd if=/dev/20010000.etf of=/cstrace.bin bs=1
>>> [ 120.009698] INFO: rcu_preempt detected stalls on CPUs/tasks:
>>> [ 120.015307] 2-...: (1 GPs behind) idle=f11/140000000000000/0
>>> softirq=224/224 fqs=1903
>>> [ 120.023226] (detected by 1, t=5255 jiffies, g=-1, c=-2, q=19)
>>> [ 120.029001] Task dump for CPU 2:
>>> [ 120.032190] dd R running task 0 1270 1267
>>> 0x00000002
>>> [ 120.039172] Call trace:
>>> [ 120.041594] [<ffff000008085534>] __switch_to+0xc8/0xd4
>>> [ 120.046675] [<0000000000020000>] 0x20000
>>>
>>> Steps followed,
>>> # git clone
>>> git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
>>> linux-next
>>> # cd linux-next
>>> # make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- defconfig
>>> # make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- menuconfig <---
>>> enable coresight
>>> # make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- -j8 Image
>>> # make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- dtbs
>>> # arch/arm64/boot/Image <--- copied this kernel
>>> # arch/arm64/boot/dts/arm/juno-r1.dtb <--- copied this dtb
>>>
>>> Top commit in linux-next is,
>>>
>>> commit c7d3b912180a9bb0733e5cfab84e5a7493dd3599
>>> Author: Stephen Rothwell <sfr at canb.auug.org.au>
>>> Date: Tue Oct 4 14:52:03 2016 +1100
>>>
>>> Add linux-next specific files for 20161004
>>
>>
>> Can't reproduce it here either.
>>
>> root at localhost:/sys/bus/coresight/devices# echo 1 > 20010000.etf/enable_sink
>> root at localhost:/sys/bus/coresight/devices# echo 1 >
>> 22140000.etm/enable_source
>> root at localhost:/sys/bus/coresight/devices# dd if=/dev/20010000.etf bs=1
>> of=/root/etr.bin
>> 65536+0 records in
>> 65536+0 records out
>> 65536 bytes (66 kB) copied, 0.227546 s, 288 kB/s
>> root at localhost:/sys/bus/coresight/devices# dd if=/dev/20010000.etf bs=1
>> of=/root/etr.bin
>> 65536+0 records in
>> 65536+0 records out
>> 65536 bytes (66 kB) copied, 0.233527 s, 281 kB/s
>> root at localhost:/sys/bus/coresight/devices# echo 0 > 20010000.etf/enable_sink
>> root at localhost:/sys/bus/coresight/devices# dd if=/dev/20010000.etf bs=1
>> of=/root/etr.bin
>> 65536+0 records in
>> 65536+0 records out
>> 65536 bytes (66 kB) copied, 0.474943 s, 138 kB/s
>>
>> FWIW, here is my firmware version :
>>
>> NOTICE: Booting Trusted Firmware
>> NOTICE: BL1: v1.1(release):e04723e21362
>> NOTICE: BL1: Built : 15:39:56, Sep 1 2015
>> NOTICE: BL1: Booting BL2
>> NOTICE: BL2: v1.1(release):e04723e21362
>> NOTICE: BL2: Built : 15:42:30, Sep 1 2015
>> NOTICE: BL1: Booting BL3-1
>> NOTICE: BL3-1: v1.1(release):604d5da6f2aa
>> NOTICE: BL3-1: Built : 14:50:36, Sep 10 2015
>> UEFI firmware (version ea31f8e built at 16:35:17 on Aug 5 2015)
>>
>
> Hang is seen while trying to dump trace _after_ disabling the ETM
> source. Is it not supposed to work?.
> It works fine, when dumped before disabling ETM source. Please find
> the log below.
>
> linaro-test [rc=0]# echo 1 > 20010000.etf/enable_sink
> linaro-test [rc=0]# echo 1 > 22140000.etm/enable_source
> [ 91.792145] coresight-tmc 20010000.etf: TMC-ETB/ETF enabled
> [ 91.797719] coresight-funnel 20040000.main-funnel: FUNNEL inport 0 enabled
> [ 91.804552] coresight-funnel 220c0000.cluster0-funnel: FUNNEL
> inport 1 enabled
> [ 91.815990] coresight-etm4x 22140000.etm: ETM tracing enabled
> linaro-test [rc=0]# dd if=/dev/20010000.etf of=/cstrace.bin bs=1
> [ 108.105492] coresight-tmc 20010000.etf: TMC read start
> [ 108.404335] coresight-tmc 20010000.etf: TMC read end
> 65536+0 records in
> 65536+0 records out
> linaro-test [rc=0]# echo 0 > 20010000.etf/enable_sink
> linaro-test [rc=0]# dd if=/dev/20010000.etf of=/cstrace.bin bs=1
> [ 125.069740] coresight-tmc 20010000.etf: TMC read start
> [ 125.184370] coresight-tmc 20010000.etf: TMC read end
> 65536+0 records in
> 65536+0 records out
> linaro-test [rc=0]# echo 0 > 22140000.etm/enable_source
> [ 140.271163] coresight-etm4x 22140000.etm: ETM tracing disabled
> [ 140.276964] coresight-funnel 220c0000.cluster0-funnel: FUNNEL
> inport 1 disabled
> [ 140.284211] coresight-funnel 20040000.main-funnel: FUNNEL inport 0 disabled
> [ 140.291128] coresight-tmc 20010000.etf: TMC-ETB/ETF disabled
> linaro-test [rc=0]# dd if=/dev/20010000.etf of=/cstrace.bin bs=1
> <---- It hangs here...
The point is once you disable the source, which eventually disables
all the components upto the sink, the data in the ETF is already captured
into a buffer, so there is no coresight hardware access involved.
Also, looks like the open(/dev/20010000.etf,..) from the dd didn't complete,
as we don't see a "TMC read start" line.
Could you please :
a) Try the patch [0] and see if it helps ?
b) If a doesn't help, after disabling the source, run any other command
instead of the trace collection. e.g,
dd if=/dev/zero of=/dev/null
and see if it hangs the system in a similar way ?
Thanks
Suzuki
[0] test patch
----8>----
diff --git a/drivers/hwtracing/coresight/coresight-tmc-etf.c b/drivers/hwtracing/coresight/coresight-tmc-etf.c
index 1ffdab8..ef72ed8 100644
--- a/drivers/hwtracing/coresight/coresight-tmc-etf.c
+++ b/drivers/hwtracing/coresight/coresight-tmc-etf.c
@@ -511,13 +511,6 @@ int tmc_read_prepare_etb(struct tmc_drvdata *drvdata)
goto out;
}
- /* There is no point in reading a TMC in HW FIFO mode */
- mode = readl_relaxed(drvdata->base + TMC_MODE);
- if (mode != TMC_MODE_CIRCULAR_BUFFER) {
- ret = -EINVAL;
- goto out;
- }
-
/* Don't interfere if operated from Perf */
if (drvdata->mode == CS_MODE_PERF) {
ret = -EINVAL;
More information about the linux-arm-kernel
mailing list