[Bug runtime/15908] New: On ARM fc19 hw_watch_addr.stp example causes kernel hang and "BUG: scheduling while atomic:" and lock up

[Bug runtime/15908] New: On ARM fc19 hw_watch_addr.stp example causes kernel hang and "BUG: scheduling while atomic:" and lock up

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=15908

            Bug ID: 15908
           Summary: On ARM fc19 hw_watch_addr.stp example causes kernel
                    hang and "BUG: scheduling while atomic:" and lock up
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: runtime
          Assignee: systemtap at sourceware dot org
          Reporter: wcohen at redhat dot com

On both Samsung arm chromebook and compulab trimslice runing fc19 running the
tests with a "make installcheck" causes the machines to hang when the
hw_watch_addr.stp test is run.  The trimslice is running an updated version of
fc19 (as of 2013-08-29) with kernel-3.10.9-200.fc19.armv7hl

chromebook has locally built linux-3.11-rc7 kernel built with the instructions
on:
http://people.redhat.com/wcohen/chromebook/booting-3.11kernel.txt

To reproduce:
-check out the current git repo of sytemtap
-build and install
-run the examples with:

make installcheck RUNTESTFLAGS="--debug systemtap.examples/check.exp"

Eventually gets to running hw_watch_addr.stp and the machine hangs.  One the
trimslice the serial console spews out a lot of output of the form:

FAIL: ./systemtap.examples/general/stopwatches build
FAIL: ./systemtap.examples/io/iostat-scsi build
[ 2332.364553] BUG: scheduling while atomic: systemd-journal/157/0x40000001
[ 2332.394534] BUG: scheduling while atomic: systemd-journal/157/0x40000001
[ 2332.394545] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.404517] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.414450] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.424526] BUG: scheduling while atomic: systemd-journal/157/0x40000001
[ 2332.424532] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.434519] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.444443] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.464532] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.474528] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.494513] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.494539] BUG: scheduling while atomic: systemd-journal/157/0x40000001
[ 2332.514516] BUG: scheduling while atomic: systemd-journal/157/0x40000001
[ 2332.514526] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.527834] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
[ 2332.544550] BUG: scheduling while atomic: rs:main Q:Reg/309/0x40000001
...

After this need to reboot the machine. Would expect that the test runs or
fails, but doesn't take the machine down.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/15908] On ARM fc19 hw_watch_addr.stp example causes kernel hang and "BUG: scheduling while atomic:" and lock up

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=15908

--- Comment #1 from William Cohen <wcohen at redhat dot com> ---
The testsuite/systemtap.log has the following output which seems to indicate
that it should fail because "OUT WARNING" says no hardware watchpoints are
available:

meta taglines 'test_support: perf stat --event=mem:0xc000000 /bin/true' tag
'test_support' value 'perf stat --event=mem:0xc000000 /bin/true'
TEST
PWD=/mnt/badlands/home/wcohen/systemtap_write/systemtap/testsuite/systemtap.examples/memory
attempting command perf stat --event=mem:0xc000000 /bin/true
OUT 
 Performance counter stats for '/bin/true':

                 0 mem:0xc000000                                               

       0.012679596 seconds time elapsed

RC 0
PASS: ./systemtap.examples/memory/hw_watch_addr support
meta taglines 'test_check: stap --all-modules -p4 hw_watch_addr.stp 0x`grep
"vm_dirty_ratio" /proc/kallsyms`' tag 'test_check' value 'stap --all-modules
-p4 hw_watch_addr.stp 0x`grep "vm_dirty_ratio" /proc/kallsyms`'
attempting command stap --all-modules -p4 hw_watch_addr.stp 0x`grep
"vm_dirty_ratio" /proc/kallsyms`
OUT WARNING: Too many hardware breakpoint probes requested for arm (1 vs. 0)
/mnt/badlands/home/wcohen/systemtap_write/systemtap/testsuite/.systemtap-root/cache/1e/stap_1ecd6f7036b3c87d801be07b2476bfa4_2556.ko
RC 0
PASS: ./systemtap.examples/memory/hw_watch_addr build
meta taglines 'test_installcheck: stap --all-modules hw_watch_addr.stp 0x`grep
"vm_dirty_ratio" /proc/kallsyms` -c "sleep 5"' tag 'test_installcheck' value
'stap --all-modules hw_watch_addr.stp 0x`grep "vm_dirty_ratio" /proc/kallsyms`
-c "sleep 5"'
attempting command stap --all-modules hw_watch_addr.stp 0x`grep
"vm_dirty_ratio" /proc/kallsyms` -c "sleep 5"

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/15908] On ARM fc19 hw_watch_addr.stp example causes kernel hang and "BUG: scheduling while atomic:" and lock up

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=15908

--- Comment #2 from William Cohen <wcohen at redhat dot com> ---
This problem can replicate more directly with:

../install/bin/stap -v -p4 -k -m hw_watch_addr --all-modules
testsuite/systemtap.examples/memory/hw_watch_addr.stp 0x`grep "vm_dirty_ratio"
/proc/kallsyms` -c "sleep 5"
 sudo ../install/bin/staprun hw_watch_addr.ko  -c "sleep 5"

-- 
You are receiving this mail because:
You are the assignee for the bug.