[igt-dev] [PATCH i-g-t] lib/kmod: reimplement kmsg_dump()

[Lucas De Marchi]

On Fri, Sep 27, 2019 at 09:54:00PM +0100, Chris Wilson wrote:
>Quoting Lucas De Marchi (2019-09-27 21:32:03)
>> On Fri, Sep 27, 2019 at 01:24:20PM -0700, Lucas De Marchi wrote:
>> >On Fri, Sep 27, 2019 at 09:09:50PM +0100, Chris Wilson wrote:
>> >>Quoting Lucas De Marchi (2019-09-20 02:07:23)
>> >>>Now the line in (1) print as:
>> >>>        (i915_selftest:5070) igt_kmod-WARNING: [drm:intel_pch_type [i915]] Found Tiger Lake LP PCH
>> >
>> >...
>> >
>> >>>
>> >>>This also fixes a double close on fclose() already closes the file
>> >>>descriptor.
>> >>>
>> >>>Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
>> >>
>> >>Too bad it doesn't actually work and bails on the first read with EPIPE.
>> >
>> >humn... I copy-pasted the line above.
>> >
>> >Checking printk.c it will happen if we explode the ringbuffer with log messages
>> >before actually reading them. You may want to increase the buffer size
>> >to get them all, but it may make sense to continue on EPIPE since the
>> >behavior is to reset to the first available message?
>>
>> btw AFAICS the previous impl would equally bail out due to getline()
>> returning -1. Or did I miss anything? (also the default buffer size used
>> by glibc is the same as the one I'm using here)
>
>It's bailing in CI (we've had no output since) and that's the first time
>I've seen it fail to provide any of the kmsg. Maybe fdopen() is
>stripping off the O_NONBLOCK?

O_NONBLOCK shouldn't make a difference. Let me send the patch below
to see if we paper it over.

I just tested this by reverting e5de91e68c5c ("Revert "drm/i915/tgl: Implement Wa_1406941453"")
and running `./build/tests/i915_selftest --run-subtest live_workarounds` on TGL and I do
get the kmsg (but I have the same behavior without this patch, do).

I see a failure in CI:
https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6971/fi-icl-dsi/igt@i915_selftest@live_hangcheck.html

My theory about a huge log seems correct.

---8<---
diff --git a/lib/igt_kmod.c b/lib/igt_kmod.c
index 01faca39..c3da4667 100644
--- a/lib/igt_kmod.c
+++ b/lib/igt_kmod.c
@@ -391,8 +391,15 @@ static void kmsg_dump(int fd)
 		if (r < 0) {
 			if (errno == EINTR)
 				continue;
+
+			if (errno == EPIPE) {
+				igt_warn("kmsg truncated: too many messages. You may want to increase log_buf_len in kmcdline\n");
+				continue;
+			}
+
 			if (errno != EAGAIN)
-				igt_warn("kmsg truncated due to unknown error: %m\n");
+				igt_warn("kmsg truncated: unknown error (%m)\n");
+
 			break;
 		}
 
---8<---