Differential D50556

PT_CONTINUE: undo transparent attach consequences
ClosedPublic
Actions

Authored by kib on May 27 2025, 9:11 PM.

Tags

None

Referenced Files

	Unknown Object (File)
	Thu, May 14, 3:30 AM

	Unknown Object (File)
	Mon, May 11, 9:38 PM

	Unknown Object (File)
	Mon, May 11, 9:01 PM

	Unknown Object (File)
	Mon, May 11, 8:38 PM

	Unknown Object (File)
	Fri, May 8, 3:34 AM

	Unknown Object (File)
	Sun, Apr 26, 11:28 AM

	Unknown Object (File)
	Thu, Apr 23, 4:42 AM

	Unknown Object (File)
	Sat, Apr 18, 1:58 AM

View All 89 Files

Subscribers

pjfloyd_wanadoo.fr

Details

Reviewers

markj
jhb

Commits

rG99976934274d: PT_CONTINUE: undo transparent attach consequences

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

kib created this revision.May 27 2025, 9:11 PM

Herald added subscribers: olce, imp. · View Herald TranscriptMay 27 2025, 9:11 PM

kib requested review of this revision.May 27 2025, 9:11 PM

emaste added a subscriber: emaste.May 27 2025, 10:33 PM

pjfloyd_wanadoo.fr added a subscriber: pjfloyd_wanadoo.fr.May 28 2025, 6:02 AM

See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=287050

I tested this with the two scenarios in the above bugzilla item.

First I had to implement sysarch AMD64_{GS}ET_TLSBASE in Valgrind as per D50482
I've done that locally, need to test it more
https://bugs.kde.org/show_bug.cgi?id=504936

With that and the patch in this differential I get no more hangs. One flaky Valgrind gdbserver test is failing and one other giving me a diff with
+unexpected result from select: No such process
That looks like a separate issue.

kib added reviewers: markj, jhb.May 28 2025, 8:49 PM

markj added inline comments.May 29 2025, 2:08 PM

sys/kern/sys_process.c
1237	If I understand correctly, this flag is always cleared immediately below. Why not use a local flag variable?

kib added inline comments.May 29 2025, 2:22 PM

sys/kern/sys_process.c
1237	It is (or should) be set by other ops too, for instance PT_SETREGS.

markj accepted this revision.May 29 2025, 2:24 PM

This revision is now accepted and ready to land.May 29 2025, 2:24 PM

Closed by commit rG99976934274d: PT_CONTINUE: undo transparent attach consequences (authored by kib). · Explain WhyMay 29 2025, 2:29 PM

This revision was automatically updated to reflect the committed changes.

kib mentioned this in rGb57eb8059665: kern_sig.c: make pt_attach_transparent global.

kib mentioned this in rGdef0c056d338: PT_CONTINUE: if pc is set, mark the thread state as modified by user.

kib added a commit: rG99976934274d: PT_CONTINUE: undo transparent attach consequences.

Revision Contents
Changeset List

Path

Size

sys/

kern/

22 lines

Diff 156253

sys/kern/sys_process.c

Loading...