Add test case for >65535 segment coredumps

[cem]

Summary

Follow-up to https://reviews.freebsd.org/D7255 .

[ngie]

tests/sys/kern/coredump_phnum_test.sh

2

Add licensing info to this file?

7

Sort by name?

17

This needs

atf_set "allow_sysctl_side_effects" "1"

34

Where does 66169 come from?

[cem]

tests/sys/kern/coredump_phnum_test.sh

2

Some other such tests don't have licensing. What's the suggested format for these?

7

Ok.

17

In head()? Ok.

34

It doesn't have any real significance. It's just the correct result of running readelf against the output of the helper program. The helper program asks for 66535 mmap segments and ends up with 66169 kernel mappings.

Well, ok, the significance is it's larger than 65535 (UINT16_MAX). The wrong (and previous) answer here is 633 (66169 mod 65536).

[cem]

Sort atf_sets and add sysctl knob.

[ngie]

tests/sys/kern/coredump_phnum_helper.c

38

main(void)
?

tests/sys/kern/coredump_phnum_test.sh

11

This violates the ATF sandbox. Stuff needs to be copied from atf_get_srcdir to the sandbox.

[ngie]

tests/sys/kern/coredump_phnum_test.sh

34

Please add a comment to this effect.

[cem]

Try to avoid violating ATF sandbox. Does it look okay now? Does the _cleanup
routine run in the same sandbox directory as the _body?

Added a comment about the magic numbers.

[emaste]

Seems fine to me but I hope one of the kyua experts can accept

[cem]

In D7264#161644, @emaste wrote:

Seems fine to me but I hope one of the kyua experts can accept

Yeah, I've been waiting for review/sign-off from someone familiar with kyua. :-)

[jmmv]

tests/sys/kern/Makefile

7

I'd rather make this BINDIR.coredump_phnum_helper and move below with the test program definition. Otherwise it's not obvious why this is here.

tests/sys/kern/coredump_phnum_helper.c

38

Add comment to describe what this helper program is intended to do please. I find the fact that this never exits successfully surprising.

tests/sys/kern/coredump_phnum_test.sh

14

Remove indentation from generated script.

16

Quote the $() expansion.

23

Are all these system-wide changes guaranteed to succeed? I think you should make sure these calls all exit successfully, and error out otherwise.

If there are legitimate cases for these calls to fail, then consider skipping the test case instead of failing it.

51

Check for the existence of the script with "test -x" before running it. The script might not have been created before the cleanup runs if the test crashes early, and you don't want to double-error on cleanup.

[cem]

tests/sys/kern/Makefile

7

I can move the line below. Does BINDIR.coredump_phnum_helper actually work?

tests/sys/kern/coredump_phnum_helper.c

38

I can add a comment. It does have "coredump" in the name… :-)

tests/sys/kern/coredump_phnum_test.sh

14

<<- does so.

16

kern.coredump is an integer.

23

They should all succeed (run by root). Is there a good way to set e in atf-sh tests? I suppose atf_check -s exit:0 would work.

51

Ok.

[cem]

Move BINDIR near coredump_phnum_helper.

Add comment describing goals of coredump_phnum_helper.

Only run cleanup helper script if it exists and is executable.

tests/sys/kern/coredump_phnum_test.sh

23

I don't think there's any reason for these to fail.