In D40772#969417, @mjg wrote:

If you can rebase both changes and show me how to collect fragmentation stats I can test this against a full ports tree build.

The summary tells the value ranges from 0 to 1. The reality is that it ranges from INT_MIN to 1000.

@alc you asked about a way to measure the fragmentation, is https://reviews.freebsd.org/D40575 maybe something which helps for this?

Should we move this to a mailinglist? This review is closed and any new changes will be a new review. So a discussion in this review looks off-topic to me.

Build failures:
lang/rust doesn't build. The same version builds just fine before the libsys+rtld changes.
cargo-c doesn't build. Same symptom than rust. Both fails in "libc" as shown above in the comment.
I don't have enough knowledge about rust to provide more info.

Can the issues this patch is supposed to fix lead to compile issues? I see failing rust and cargo-c builds in the "libc" part. And I see issues with nslookup to a bind 9.18 which serves an internal domain.

I updated the src but build only ld-elf.so.1. What is committed works. I will do a full buildworld now. Consider it OK unless I come back to you. ;-)

In D43858#1000998, @krion wrote:

hm, I applied this one https://people.freebsd.org/~krion/D43858.diff and it worked

In D43858#1001000, @kib wrote:

In D43858#1000997, @netchild wrote:

In D43858#1000995, @krion wrote:

@netchild are you sure you applied the correct patch?

This is what I use:

It is hard to read your paste, and I cannot find there the critical chunk from https://reviews.freebsd.org/rG968a18975adc9c2a619bb52aa2f009de99fc9e24.

In D43858#1000996, @kib wrote:

In D43858#1000992, @netchild wrote:

Mittwoch, 14. Februar 2024, 10:47:15
{0} <FreeBSD 15.0-CURRENT> [imap:/]
(210) root@ttypts/2 # ll /var/run/dovecot/*.core
zsh: no matches found: /var/run/dovecot/*.core

Mittwoch, 14. Februar 2024, 10:47:19
{1} <FreeBSD 15.0-CURRENT> [imap:/]
(211) root@ttypts/2 # ll /var/run/dovecot/*.core
-rw------- 1 root wheel 244M 14 Feb. 10:47 /var/run/dovecot/imap.core

Mittwoch, 14. Februar 2024, 10:47:37
{0} <FreeBSD 15.0-CURRENT> [imap:/]
(211) root@ttypts/2 # lldb --core /var/run/dovecot/imap.core --batch -o 'thread backtrace all' -o 'quit' /usr/local/libexec/dovecot/imap >/tmp/ld-elf.core.txt

Mittwoch, 14. Februar 2024, 10:55:03
{0} <FreeBSD 15.0-CURRENT> [imap:/]
(212) root@ttypts/2 # head /tmp/ld-elf.core.txt
(lldb) target create "/usr/local/libexec/dovecot/imap" --core "/var/run/dovecot/imap.core"
Core file '/var/run/dovecot/imap.core' (x86_64) was loaded.
(lldb) thread backtrace all

• thread #1, name = 'imap', stop reason = signal SIGSEGV
  • frame #0: 0x000022518aadc5e1 ld-elf.so.1`load_object [inlined] object_match_name(obj=0x000009030f403408, name="") at rtld.c:5613:6 frame #1: 0x000022518aadc5c2 ld-elf.so.1`load_object(name="", fd_u=-1, refobj=0x000009030f428008, flags=0) at rtld.c:2704:10 frame #2: 0x000022518aadbd2a ld-elf.so.1`dlopen_object(name="", fd=-1, refobj=0x000009030f428008, lo_flags=0, mode=1, lockstate=0x000001edfeaa0f30) at rtld.c:3747:8 frame #3: 0x000022518aada0b4 ld-elf.so.1`symlook_obj_load_filtees [inlined] load_filtee1(obj=0x000009030f428008, needed=0x000009030f4007c8, flags=0, lockstate=0x000001edfeaa0f30) at rtld.c:2576:16 frame #4: 0x000022518aada096 ld-elf.so.1`symlook_obj_load_filtees [inlined] load_filtees(obj=0x000009030f428008, flags=0, lockstate=0x000001edfeaa0f30) at rtld.c:2589:2 frame #5: 0x000022518aada01c ld-elf.so.1`symlook_obj_load_filtees(req=0x000001edf0264070, req1=0x000001edf0263ff8, obj=0x000009030f428008, needed=0x000009030f4007c8) at rtld.c:4701:2

How this excerpt from the backtrace answers my query? I want to see the content of the object located at the address I provided.

In D43858#1000995, @krion wrote:

@netchild are you sure you applied the correct patch?

nm /libexec/ld-elf.so.1 | grep 0x00004da

Mittwoch, 14. Februar 2024, 10:47:15
{0} <FreeBSD 15.0-CURRENT> [imap:/]
(210) root@ttypts/2 # ll /var/run/dovecot/*.core
zsh: no matches found: /var/run/dovecot/*.core

nm /libexec/ld-elf.so.1 | grep symlook_obj_load_filtees

000000000000cfe0 t symlook_obj_load_filtees

For the benefit of subscribers: the biiiiiig backtrace has been provided out-of-band.

Dovecot crashes with this patch. With first patch via email it works for me.

This is the "install" part only. Is this review also supposed to get the dependency part for this to work (e.g. a port depending on only a subpackage and the recursive dependency tracking and install needs to set TARGET_SUBPACKAGE), or is the dependency part supposed to come / available in another review?

The comments for rc.subr also apply to the new files.
What about a man page update how to turn on the debug feature? To my current understanding DEBUG_SH can be set interactively in the loader, and administratively in loader.conf... maybe also in rc.conf? So more than one man page to touch at least with a cross-ref.

I have reworded the rc-scripting part. Does this make it more easy / clear?

Address comments in the review.

• make some scripts compatible with svcj (convert parts of the precmd into another way of settings variables), precmd is not run inside the same shell/jail = make it work with svcj
• add some support for nfs in svcj, not yet finished (precmd is not comaptible)
• exclude some scripts from svcj due to an incompatible precmd (not run in same shell/jail)

Make jls quiet.

Address issues from comments.

Only rc.d/opensm is missing.

Add support for nfs. Sort the options.

Add config for some more services.

Change what was noticed in comments. Add a feature to enable the execution of extra commands inside the service jail.

Fix variable names.