Retry_count in minidump 64 bit code delaying ENOSPC by around 30 minutes for live cores.

[raviprakash.darbha_gmail.com]

Summary

I notice that when, the system runs out of space and we are trying to get a live core dump there is delay of around 30 minutes to return ENOSPC to the upper layers. This is on AMD64. As I go through the code , I see a retry count in amd64 not present in other platforms in sys/amd64/amd64/minidump_machdep.c . Its compared against 5 ( Not sure where that number comes from ). Because of this, the live core generation in our platform is delayed and as the memory map keeps increasing( lots of other stuff running ) it makes things worse. I want to make this retry count a tunable so that it can be changed at runtime accordingly.

Test Plan

I have verified this behavior on our platform and works as intended.

[cem]

Please upload a diff with full context. (Either use diff -U9999 or arc.)

sys/amd64/amd64/minidump_machdep.c

77–78 ↗

(On Diff #21404)

This can just be SYSCTL_INT(..., CTLFLAG_RWTUN, ...).

[raviprakash.darbha_gmail.com]

added the full file for more context. Also merged TUNABLE and SYSCTL_INT into one command.

[markj]

LGTM with those nits fixed.

sys/amd64/amd64/minidump_machdep.c

35 ↗

(On Diff #21424)

This should be sorted between msgbuf.h and watchdog.h.

72 ↗

(On Diff #21424)

dump_retry_count seems like a better name to me.

[raviprakash.darbha_gmail.com]

Addressed comments by markj