9:10 a.m.
The sys command displays essential system information and it is often
shown by default when crash is started in non-silent mode. It might be
considered helpful to report if the kernel is tainted or not when
sys is run without any arguments. This patch makes this change.
The intended output is as follows:
KERNEL: /usr/lib/debug/lib/modules/3.16.4-200.fc20.x86_64/vmlinux
DUMPFILE: /dev/crash
CPUS: 4
DATE: Fri Nov 21 15:02:56 2014
UPTIME: 6 days, 07:41:22
LOAD AVERAGE: 0.29, 0.20, 0.15
TASKS: 397
NODENAME: atomlin.usersys.redhat.com
RELEASE: 3.16.4-200.fc20.x86_64
VERSION: #1 SMP Mon Oct 6 12:57:00 UTC 2014
TAINTED: YES
MACHINE: x86_64 (2693 Mhz)
MEMORY: 7.7 GB
PID: 12172
COMMAND: "crash"
TASK: ffff8801b0aacf00 [THREAD_INFO: ffff8801b1e1c000]
CPU: 0
STATE: TASK_RUNNING (ACTIVE)
If the tainted_mask or tainted symbol does not exist then nothing is
displayed.
Signed-off-by: Aaron Tomlin <atomlin(a)redhat.com>
---
kernel.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/kernel.c b/kernel.c
index 1cb0967..a22ce60 100644
--- a/kernel.c
+++ b/kernel.c
@@ -4691,8 +4691,12 @@ display_sys_stats(void)
{
struct new_utsname *uts;
char buf[BUFSIZE];
+ ulong tainted_mask;
+ int tainted;
ulong mhz;
+ tainted_mask = tainted = 0;
+
uts = &kt->utsname;
// if (!(pc->flags & RUNTIME) && !DUMPFILE() && !GDB_PATCHED())
@@ -4834,6 +4838,13 @@ display_sys_stats(void)
fprintf(fp, " NODENAME: %s\n", uts->nodename);
fprintf(fp, " RELEASE: %s\n", uts->release);
fprintf(fp, " VERSION: %s\n", uts->version);
+ if (kernel_symbol_exists("tainted_mask")) {
+ get_symbol_data("tainted_mask", sizeof(ulong), &tainted_mask);
+ fprintf(fp, " TAINTED: %s\n", tainted_mask ? "YES" :
"NO");
+ } else if (kernel_symbol_exists("tainted")) {
+ get_symbol_data("tainted", sizeof(int), &tainted);
+ fprintf(fp, " TAINTED: %s\n", tainted ? "YES" :
"NO");
+ }
fprintf(fp, " MACHINE: %s ", uts->machine);
if ((mhz = machdep->processor_speed()))
fprintf(fp, "(%ld Mhz)\n", mhz);
--
1.9.3
9:25 a.m.
New subject: [PATCH] crash: Show kernel tainted status in display_sys_stats()
Kind of redundant w/respect to the "sys -t" option that was added
in crash-7.0.9?
Dave
----- Original Message -----
The sys command displays essential system information and it is
often
shown by default when crash is started in non-silent mode. It might be
considered helpful to report if the kernel is tainted or not when
sys is run without any arguments. This patch makes this change.
The intended output is as follows:
KERNEL: /usr/lib/debug/lib/modules/3.16.4-200.fc20.x86_64/vmlinux
DUMPFILE: /dev/crash
CPUS: 4
DATE: Fri Nov 21 15:02:56 2014
UPTIME: 6 days, 07:41:22
LOAD AVERAGE: 0.29, 0.20, 0.15
TASKS: 397
NODENAME: atomlin.usersys.redhat.com
RELEASE: 3.16.4-200.fc20.x86_64
VERSION: #1 SMP Mon Oct 6 12:57:00 UTC 2014
TAINTED: YES
MACHINE: x86_64 (2693 Mhz)
MEMORY: 7.7 GB
PID: 12172
COMMAND: "crash"
TASK: ffff8801b0aacf00 [THREAD_INFO: ffff8801b1e1c000]
CPU: 0
STATE: TASK_RUNNING (ACTIVE)
If the tainted_mask or tainted symbol does not exist then nothing is
displayed.
Signed-off-by: Aaron Tomlin <atomlin(a)redhat.com>
---
kernel.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/kernel.c b/kernel.c
index 1cb0967..a22ce60 100644
--- a/kernel.c
+++ b/kernel.c
@@ -4691,8 +4691,12 @@ display_sys_stats(void)
{
struct new_utsname *uts;
char buf[BUFSIZE];
+ ulong tainted_mask;
+ int tainted;
ulong mhz;
+ tainted_mask = tainted = 0;
+
uts = &kt->utsname;
// if (!(pc->flags & RUNTIME) && !DUMPFILE() && !GDB_PATCHED())
@@ -4834,6 +4838,13 @@ display_sys_stats(void)
fprintf(fp, " NODENAME: %s\n", uts->nodename);
fprintf(fp, " RELEASE: %s\n", uts->release);
fprintf(fp, " VERSION: %s\n", uts->version);
+ if (kernel_symbol_exists("tainted_mask")) {
+ get_symbol_data("tainted_mask", sizeof(ulong), &tainted_mask);
+ fprintf(fp, " TAINTED: %s\n", tainted_mask ? "YES" :
"NO");
+ } else if (kernel_symbol_exists("tainted")) {
+ get_symbol_data("tainted", sizeof(int), &tainted);
+ fprintf(fp, " TAINTED: %s\n", tainted ? "YES" :
"NO");
+ }
fprintf(fp, " MACHINE: %s ", uts->machine);
if ((mhz = machdep->processor_speed()))
fprintf(fp, "(%ld Mhz)\n", mhz);
--
1.9.3
--
Crash-utility mailing list
Crash-utility(a)redhat.com
https://www.redhat.com/mailman/listinfo/crash-utility
9:48 a.m.
On Fri 2014-11-21 10:25 -0500, Dave Anderson wrote:
Kind of redundant w/respect to the "sys -t" option that was added
in crash-7.0.9?
I noticed that but initially thought it would still be nice to provide a
brief indication at start up anyway. While 'sys -t' gives us more
information much like 'mod -t' for modules I suppose.
--
Aaron Tomlin
1:12 p.m.
On Fri 2014-11-21 15:48 +0000, Aaron Tomlin wrote:
On Fri 2014-11-21 10:25 -0500, Dave Anderson wrote:
>
> Kind of redundant w/respect to the "sys -t" option that was added
> in crash-7.0.9?
I noticed that but initially thought it would still be nice to provide a
brief indication at start up anyway. While 'sys -t' gives us more
information much like 'mod -t' for modules I suppose.
Dave,
Would you still consider this for the next release? I think it would be
a nice first glance "YES/NO" indicator.
--
Aaron Tomlin
1:39 p.m.
----- Original Message -----
On Fri 2014-11-21 15:48 +0000, Aaron Tomlin wrote:
> On Fri 2014-11-21 10:25 -0500, Dave Anderson wrote:
> >
> > Kind of redundant w/respect to the "sys -t" option that was added
> > in crash-7.0.9?
>
> I noticed that but initially thought it would still be nice to provide a
> brief indication at start up anyway. While 'sys -t' gives us more
> information much like 'mod -t' for modules I suppose.
Dave,
Would you still consider this for the next release? I think it would be
a nice first glance "YES/NO" indicator.
Before I went with adding "sys -t", I was thinking of appending
"[TAINTED]" to
the DUMPFILE: line. That would make it even more noticable than adding your
extra TAINTED: line.
8:33 a.m.
----- Original Message -----
On Fri 2014-11-21 10:25 -0500, Dave Anderson wrote:
>
> Kind of redundant w/respect to the "sys -t" option that was added
> in crash-7.0.9?
I noticed that but initially thought it would still be nice to provide a
brief indication at start up anyway. While 'sys -t' gives us more
information much like 'mod -t' for modules I suppose.
--
Aaron Tomlin
I did a quick check of this patch with my set of sample dumps, and since
a majority (presumably) of vmcores are generated as the result of an oops,
they will be marked as tainted. So printing the tainted information is
pretty much useless, and actually somewhat misleading.
That's because the kernel will be marked as tainted because TAINT_DIE gets
set by oops_end(), which is called *after* __die(), which is where the
registers and the translation of the tainted mask are dumped:
die()
__die()
show_regs()
show_regs_print_info()
dump_stack_print_info()
print_tainted() --> translates tained_mask
...
oops_end() --> add_taint(TAINT_DIE)
panic()
So the kernel log will (correctly) show "Not tainted" because oops_end() would
not have been called yet, yet this patch would yell out TAINTED.
So I'd like to keep this relegated to "sys -t".
Dave
8:47 a.m.
----- Original Message -----
On 03/12/14 14:33, Dave Anderson wrote:
> So the kernel log will (correctly) show "Not tainted" because oops_end()
> would
> not have been called yet, yet this patch would yell out TAINTED.
So ignore that "die" bit wrt. the tainting for this purpose?
Suppose that there had been several prior die() calls, and then
the user decided to set "panic_on_oops".
Dave
9:15 a.m.
On Wed 2014-12-03 09:33 -0500, Dave Anderson wrote:
I did a quick check of this patch with my set of sample dumps, and
since
a majority (presumably) of vmcores are generated as the result of an oops,
they will be marked as tainted. So printing the tainted information is
pretty much useless, and actually somewhat misleading.
That's because the kernel will be marked as tainted because TAINT_DIE gets
set by oops_end(), which is called *after* __die(), which is where the
registers and the translation of the tainted mask are dumped:
die()
__die()
show_regs()
show_regs_print_info()
dump_stack_print_info()
print_tainted() --> translates tained_mask
...
oops_end() --> add_taint(TAINT_DIE)
panic()
So the kernel log will (correctly) show "Not tainted" because oops_end() would
not have been called yet, yet this patch would yell out TAINTED.
So I'd like to keep this relegated to "sys -t".
OK - understood.
--
Aaron Tomlin
3496
days inactive
3508
days old
devel@lists.crash-utility.osci.io
8 comments
3 participants
participants (3)
-
Aaron Tomlin -
Dave Anderson -
Jeremy Harris