Hi Alexey,
Thank you for the huge work! and sorry for the delayed response.
Apparently updating gdb had been done only when kernel required it,
but these days gdb-7.6 has some problems and is old enough, this
update can make crash more useful and future proof.
I need more time to finish reviewing the patches, but there are
a few lacks of output from my test. Is it possible to fix the
following?
--- crash-master.log
+++ crash-gdb-10.1.log
* 'bt' command often emits reduced output:
@@ -3391,13 +2885,6 @@
PID: 951 TASK: ffff880079489fa0 CPU: 0 COMMAND: "sshd"
#0 [ffff88007997b888] __schedule at ffffffff816a8f45
#1 [ffff88007997b8f0] schedule at ffffffff816a94c9
- #2 [ffff88007997b900] schedule_hrtimeout_range_clock at ffffffff816a86dd
- #3 [ffff88007997b998] schedule_hrtimeout_range at ffffffff816a8713
- #4 [ffff88007997b9a8] poll_schedule_timeout at ffffffff81215eb5
- #5 [ffff88007997b9d8] do_select at ffffffff81216831
- #6 [ffff88007997bd80] core_sys_select at ffffffff81216afb
- #7 [ffff88007997bf18] sys_select at ffffffff81216cda
- #8 [ffff88007997bf80] system_call_fastpath at ffffffff816b4fc9
RIP: 00007f04a6bd4783 RSP: 00007fffd1652898 RFLAGS: 00000246
RAX: 0000000000000017 RBX: ffffffff816b4fc9 RCX: ffffffffffffffff
RDX: 0000000000000000 RSI: 000055d9679059d0 RDI: 0000000000000007
* lack of information about struct members:
@@ -23937,666 +22908,538 @@
crash> task
PID: 11077 TASK: ffff88005fd59fa0 CPU: 1 COMMAND: "bash"
struct task_struct {
- state = 0,
- stack = 0xffff880062ab0000,
+ state = 0,
+ stack = , <<--
usage = {
counter = 2
...
vtime_seqlock = {
seqcount = {
sequence = 0
- },
+ },
lock = {
- {
- rlock = {
- raw_lock = {
- val = {
- counter = 0
- }
- }
- }
- }
+ {...} <<-- Can we unshorten this abbreviation? (unionprint?)
}
-#WARNING_OPTIONS=-Wall -O2 -Wstrict-prototypes -Wmissing-prototypes
-fstack-protector -Wformat-security
+WARNING_OPTIONS=-Wall -O2 -Wstrict-prototypes -Wmissing-prototypes -fstack-protector
-Wformat-security
Please don't change the default options in this patch.
Thanks,
Kazu