Bug 210557 - sys/geom broken kyua status for 11.0 -r302180 (only tested on rpi2 with -mcpu=cortex-a7 in use)
Summary: sys/geom broken kyua status for 11.0 -r302180 (only tested on rpi2 with -mcpu...
Status: New
Alias: None
Product: Base System
Classification: Unclassified
Component: tests (show other bugs)
Version: CURRENT
Hardware: arm Any
: --- Affects Some People
Assignee: freebsd-testing (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-06-25 08:59 UTC by Mark Millard
Modified: 2017-11-07 16:04 UTC (History)
3 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Mark Millard 2016-06-25 08:59:49 UTC
[The ". . ."s are omitted lines.]
(only tested on rpi2 with -mcpu=cortex-a7 in use)

kyua report --results-filter broken --results-file /usr/tests --verbose

reports for sys/geom:

===> sys/geom/class/eli/integrity_copy_test:main
Result: broken: Test case timed out
Duration: 1200.089s
 
Metadata:
    allowed_architectures is empty
    allowed_platforms is empty
    description is empty
    has_cleanup = false
    required_configs is empty
    required_disk_space = 0
    required_files is empty
    required_memory = 0
    required_programs is empty
    required_user = root
    timeout = 1200
 
Standard output:
1..5520
ok 1 - small 1 aalgo=hmac/md5 ealgo=aes keylen=0 sec=512
. . .
ok 4442 - small 2 aalgo=hmac/md5 ealgo=blowfish-cbc keylen=448 sec=512
ok 4443 - big 1 aalgo=hmac/md5 ealgo=blowfish-cbc keylen=448 sec=512
ok 4444 - big 2 aalgo=hmac/md5 ealgo=blowfish-cbc keylen=448 sec=512
 
Standard error:
Subprocess timed out; sending KILL signal...
===> sys/geom/class/eli/onetime_a_test:main
Result: broken: Test case timed out
Duration: 600.041s
 
Metadata:
    allowed_architectures is empty
    allowed_platforms is empty
    description is empty
    has_cleanup = false
    required_configs is empty
    required_disk_space = 0
    required_files is empty
    required_memory = 0
    required_programs is empty
    required_user = root
    timeout = 600
 
Standard output:
1..1380
ok 1 - aalgo=hmac/md5 ealgo=aes keylen=0 sec=512
. . .
ok 794 - aalgo=hmac/ripemd160 ealgo=blowfish-cbc keylen=0 sec=4096
 
Standard error:
Subprocess timed out; sending KILL signal...
Comment 1 Mark Millard 2016-07-05 17:42:53 UTC
(In reply to Mark Millard from comment #0)

A 11.0 -r302331 test had two more sys/geom tests timeout (so they can sometimes occur):

sys/geom/class/eli/integrity_data_test:main  ->  broken: Test case timed out  [600.142s]
sys/geom/class/eli/integrity_hmac_test:main  ->  broken: Test case timed out  [600.100s]
Comment 2 Enji Cooper freebsd_committer 2016-07-05 17:44:19 UTC
What was the frequency on the rpi2's processor again?
Comment 3 Mark Millard 2016-07-05 18:00:33 UTC
(In reply to Ngie Cooper from comment #2)

This is an Raspberry Pi 2B context. I'm not sure what FreeBSD is doing for frequency control.

As it sits at the moment it reports:

hw.cpufreq.turbo: 1
hw.cpufreq.sdram_freq: 450000000
hw.cpufreq.core_freq: 250000000
hw.cpufreq.arm_freq: 900000000
. . .
dev.bcm2835_cpufreq.0.freq_settings: 900/-1 600/-1
dev.bcm2835_cpufreq.0.%parent: cpu0
dev.bcm2835_cpufreq.0.%pnpinfo: 
dev.bcm2835_cpufreq.0.%location: 
dev.bcm2835_cpufreq.0.%driver: bcm2835_cpufreq
dev.bcm2835_cpufreq.0.%desc: CPU Frequency Control
dev.bcm2835_cpufreq.%parent: 
dev.cpu.0.freq_levels: 900/-1 600/-1
dev.cpu.0.freq: 900

I will note that I have the root file system on a USB SSD, although the kernel is from the SD card. The USB SSD combination is operationally much faster than the SD card. If there was significant I/O to the SD card things would take even longer.

While technically the USB SSD is a fast USB3 SSD on a USB3 capable hub the rpi2 only has USB2 and does not fully utilize USB2 capacity form what I've seen. (But I've done no formal benchmarks so far.)
Comment 4 Mark Millard 2016-07-05 22:43:04 UTC
(In reply to Ngie Cooper from comment #2)

For the USB SSD that I normally use as the root file system on the rpi2 here is what benchmarks/bonnie reports:

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
         4096 12024 99.8 23417 23.4  8551 10.9 12825 100.0 16128 11.9 522.9  7.9

which is for "/dev/ufs/RPI2rootfs on /" below:

# mount
/dev/ufs/RPI2rootfs on / (ufs, NFS exported, local, noatime, soft-updates)
devfs on /dev (devfs, local)
/dev/mmcsd0s1 on /boot/msdos (msdosfs, local, noatime)
/dev/ufs/RPI2Arootfs on /mnt (ufs, local, noatime, soft-updates)

# df -m
Filesystem           1M-blocks  Used  Avail Capacity  Mounted on
/dev/ufs/RPI2rootfs     440365 40251 364885    10%    /
devfs                        0     0      0   100%    /dev
/dev/mmcsd0s1               49     7     42    15%    /boot/msdos
/dev/ufs/RPI2Arootfs     26763 13215  11406    54%    /mnt

I will report what bonnie reports about the SD card after it finishes ("/dev/ufs/RPI2Arootfs on /mnt").
Comment 5 Mark Millard 2016-07-05 23:23:03 UTC
(In reply to Mark Millard from comment #1)

For the SD card that I normally use boot the kernel on the rpi2 here is what benchmarks/bonnie reports:

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
         4096  7383 60.2  9573  7.8  5896  5.9 13077 100.0 17884  8.8 457.6  5.3

It is the write speeds that are notably slower than when using the USB SSD.
Comment 6 Enji Cooper freebsd_committer 2017-11-05 21:03:30 UTC
The timeout would likely need to be increased for a number of long-running tests, at the cost of failing sooner with faster running platforms.

Not having a means of dynamically detecting platform performance baked in to ATF/Kyua is the first item that probably should be tackled.

I have no interest in doing this though, because upstreaming changes to ATF/Kyua is annoyingly bureaucratic.
Comment 7 Brad Davis freebsd_committer 2017-11-07 16:04:42 UTC
(In reply to Ngie Cooper from comment #6)

Is it worth just forking it?