[simonbronk]

What

sceNetRecvfrom opens the receive-data buffer for host writes but writes the sender address and its length without those guards. A logically read-only sender-address mapping on Windows causes an unhandled 0xc0000005 write in libSceNet.prx instead of EFAULT.

Guard the length output before the unbound-socket early return, and the datagram address before receiving. Bound the address guard by caller capacity and native family size (16 bytes for IPv4, 28 for IPv6). Zero-capacity addresses need no write permission; stream address bytes remain untouched. Refused outputs leave the queued datagram available.

The existing watched-socket fixture now covers read-only address and length outputs, unchanged metadata and packet recovery, zero address capacity, and an IPv4 output ending at a writable page followed by a read-only page despite a larger supplied capacity.

No affected title is confirmed: this is a synthetic native regression. #5930 concerns sceNetRecvmsg message/vector guards, not these sceNetRecvfrom sender outputs. The current open #5836 and #6926 do not add these guards. No new PR implementing them was found.

Tested

Windows x64, Intel Core i7-8700, supported WinLibs GCC 15.2.0 r7 UCRT, CMake 4.4.4, Release, BUILD_TESTING=ON, bundled dependencies.

Refreshed on 2026-10-17 to main 84b0d2f24863d1e88105ada295cc346f720c1a6a. Exact tested tree ba773e46de2fa81663009f4fca80f348201c3b64. The rebase preserves main's new unbound-socket wait; the length guard also covers that path's length write.

• Rebuilt the actual network library, runtime dependencies and guest_socket_host_write_tests, guest_datagram_receive_tests, guest_sce_net_tests on this base.
• Final related network CTests and ten repetitions of the extended UDP regression pass with no skips; raw output attached.
• Negative control captured on the original base 82ddae75b6761757f3d3ba9c0723647469270f76: adding the read-only-address case without production changes causes 0xc0000005 at the sender-address write. The original fixture without that case passed there. This negative control was not repeated on the new base.
• Conventions against the new base and git diff --check pass. pr_overlap.py could not run because gh is unavailable; overlap was checked through the GitHub connector and the relevant open PR diffs.
• New protected-output cases are IPv4 UDP. Existing TCP/IPv6 normal-path controls pass. No protected IPv6-output test, full all-target build, full CTest sweep, Linux/macOS or title execution. No performance claim.

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 under Silent stubs in their library's technical debt file
• 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
• Fixes: none
• Unblocks: none (no affected title verified)
• GPUs tested: n/a
• AI-assisted: yes (OpenAI Codex)
  net-receive-address-evidence-20261017.zip

[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
PR 6044	core/libs/tests/GuestSocketHostWrite.cpp
PR 6447	core/libs/prx/libSceNet/Export.cpp
PR 7201	core/libs/prx/libSceNet/Export.cpp
PR 7332	core/libs/prx/libSceNet/Export.cpp