[milescoviello]

What

MXGP 26: The Official Game (PPSA30839) opens TCP sockets with libkernel's socket(), then closes them with sceNetSocketClose. sceNetSocketClose only looked in libSceNet's own socket table, so it failed every one of those closes with SCE_NET_ERROR_EBADF (0x80410109), and the host socket stayed open. On the console, libSceNet's calls and libkernel's socket calls work on the same kernel sockets.

Now, when sceNetSocketClose gets a descriptor that libSceNet doesn't own but libkernel has open as a socket (GuestSocketIsOpen_nid_no_patch), it closes it through libkernel's close(). Any other descriptor still fails with EBADF, as before.

The fallback runs after g_mutex is released. libkernel's close() first asks libSceNet's registered socket closer, and that closer takes g_mutex.

guest_sce_net gains a check:

• it opens a TCP socket with libkernel's socket() and closes it with sceNetSocketClose;
• libkernel no longer reports the socket open;
• a second sceNetSocketClose fails with EBADF.

Tested

Ubuntu 24.04, GCC 13.3 (CI's compiler), Release with CI's configure flags, on main eea428c:

• guest_sce_net aborts with main's Export.cpp and passes with the change.
• All 376 guest_* tests pass.
• tools/check_conventions.py --base upstream/main passes.
• PPSA30839 relinked with --to-intel, 90 s, every libSceNet socket call traced under gdb:
  • On main, sceNetSocketClose returns 0x80410109 for descriptors 7, 9 and 11. 7 and 9 came from libkernel's socket(2, 1, 0).
  • With this change, all three return 0.
  • Its sceNetSetsockopt(11, IPPROTO_TCP, TCP_NODELAY) still fails with EBADF. Setting options on a libkernel socket through libSceNet is a separate change, and feat(libkernel): answer TCP_NODELAY, SO_NOSIGPIPE and SO_NBIO on a guest socket #5565 covers TCP_NODELAY on libkernel sockets.

Not checked on a console.

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 (NotImplemented_nid_no_patch); silent stubs are listed in TechnicalDebt
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• AI-assisted: yes

🤖 Generated with Claude Code

[github-actions]

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 5836	core/libs/prx/libSceNet/Export.cpp core/libs/tests/GuestSceNet.cpp
PR 6044	core/libs/tests/GuestSceNet.cpp
PR 6138	core/libs/tests/GuestSceNet.cpp
PR 7201	core/libs/prx/libSceNet/Export.cpp

[adriadam10]

Checked build and ctest (guest_sce_net passes; it fails when Export.cpp is reverted), the lock ordering (fallback runs after g_mutex is released) and the EBADF path for descriptors neither library owns; merging.

[milescoviello]

@adriadam10, I rebased this onto main e383833 earlier today, as part of rebasing my older PRs. The head moved from 1c65051 to 8b81fa0 and the patch is unchanged (git range-diff shows =). In case your merge didn't go through: after the rebase, guest_sce_net passes 3 of 3 and ^guest_ passes 409 of 409.

AI-assisted: yes (Claude Code)

[oneandonlydean]

Merged while @adriadam10 is away: the review passed and full CI is green on this head.