[xXPBSXx]

What

AddressLengthError (the length and null checks of Address(), i.e. FreeBSD getsockaddr) now runs before Lookup() in bind, connect and sendto (only when an address is given). Valid arguments on a closed descriptor still fail with EBADF; the family and short-AF_INET checks stay after the descriptor lookup.
Fixes #8391
Follow-up to #8154 (socket address length). setsockopt is not changed here; the pointer and length checks of setsockopt/getsockopt are in #7511. git merge-tree against the open socket PRs #7490, #7850, #7475 and #8312 is clean; pr_overlap was not run.

Tested

Windows 10 (19045), MinGW-w64 GCC 15.2.0, Release, on upstream/main d50ec76 + this commit: guest_socket_address_length fails without the change to Udp.cpp (bind(-1) with a 1-byte address answers 9) and passes with it; 28 other socket and sceNet tests pass (29 of 29); python tools/check_conventions.py --base upstream/main passes. Not tested: Linux/macOS, a console, a title run.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw; no silent stubs added
• No notes, investigation .md files or images added to the repository
• New third-party code is a submodule built from source (none added)
• Generic behaviour, not specific to one title
• Depends on: none
• Unblocks: none
• AI-assisted: yes

AI-assisted (Claude Code).

🤖 Generated with Claude Code

[github-actions]

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Pull request	Conflicting files (lines on main)	Same exports	Same new files
PR 9169	core/libs/tests/GuestSocketAddressLength.cpp L58, L70

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 5021	core/libs/prx/libkernel/Socket/src/Udp.cpp
PR 7201	core/libs/prx/libkernel/Socket/src/Udp.cpp

[xXPBSXx]

Overlap note: #7511 (validate the pointers of getsockopt and setsockopt like FreeBSD) adds the same null-value and negative-length checks for setsockopt. git merge-tree of the two heads is clean, but the setsockopt part of this PR then repeats those checks; this PR only moves them in front of the descriptor lookup, which is what kern_setsockopt does (uipc_syscalls.c:1414-1417). If #7511 lands first, I can rebase this to the descriptor-order change only. I should have listed #7511 in the description; pr_overlap was not run before opening.

[oneandonlydean]

Thanks, the ordering matches FreeBSD's getsockaddr before getsock. One thing in the description: it says setsockopt now rejects a null value with a non-zero length and a negative length before the descriptor, but the diff doesn't change setsockopt, and it still looks the descriptor up first. Could you either drop that sentence or add the change with a test?

[xXPBSXx]

Thanks, you are right: the diff changes only the address checks of bind, connect and sendto; setsockopt still looks the descriptor up first. The setsockopt sentence came from an earlier version of this change that I dropped because #7511 covers the pointer and length checks of setsockopt/getsockopt. I removed it from the description, the title and the linked issue. My earlier comment here about the setsockopt part overlapping #7511 was therefore also wrong: there is no overlap in setsockopt.

[xXPBSXx]

Linux data point for ce8f7803ea (WSL2, Ubuntu 24.04, GCC 13.3, Release, -j4; a virtualised kernel, not bare-metal Linux): the socket and UDP tests pass on Linux (14 of 14). This replaces the "Linux not tested" line in the description for this head. Not tested: macOS.

AI-assisted (Claude Code).

[adriadam10]

Checked the order of bind, connect and sendto against FreeBSD 11.4 (getsockaddr before getsock), the new test pins the length/null errors before EBADF and fails on main, the family checks stay after the lookup, and Windows plus WSL2 runs are in the thread; merged.