On macOS, shutdown(SHUT_RD) on a UDP socket that was never bound does not
wake a thread blocked in recv on it, so sceNetShutdown left the receiver
hanging. Use the receive wait that Windows already has: block on the
socket table until the socket gets a local port or is shut, then poll.
On macOS an unbound socket reports port 0 from getsockname, so the check
looks at the port instead of the Winsock error.
Ported from #8658.
Co-authored-by: Himansh Raj <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
sceNetConfigGetDefaultRoute threw. It now writes the IPv4 gateway of the
host interface that libSceNetCtl's sceNetCtlGetInfo already describes as
DEFAULT_ROUTE, in network byte order, the order sceNetInetPton writes an
AF_INET address in, and returns 0. libSceNetCtl exports
NetCtlHostIpv4Gateway_nid_no_patch for it, next to the DNS helper
sceNetGetDnsInfo uses. Without a gateway, or for a null pointer, it
still throws: the console's answer for a missing route is not measured.
Tested: guest_sce_net_tests, after sceNetCtlInit, prints the gateway
with sceNetInetNtop and compares it with sceNetCtlGetInfo's
DEFAULT_ROUTE text (192.0.2.1 on this host); a null pointer throws.
On a host without a default route the case checks that the call
throws and leaves the output untouched.
sceNetInetNtopWithScopeId and sceNetInetPtonWithScopeId threw. They now
handle AF_INET6 text with an RFC 4007 numeric zone (fe80::1%2): the
address part goes through sceNetInetNtop and sceNetInetPton, so it
follows the FreeBSD inet text rules, and the zone is printed and parsed
as FreeBSD 11.4's getnameinfo (NI_NUMERICSCOPE) and getaddrinfo do. A
zero scope ID prints the address alone and text without a zone parses
to scope ID 0. An empty zone, one above 2^32 - 1, or an invalid address
part returns 0 and leaves both outputs untouched; a buffer too small
for the scoped text gives ENOSPC.
A zone given as an interface name, and AF_INET, still throw: the tree
has no mapping from interface names to the console's interface indexes
and no answer for an IPv4 zone. The debt file says so.
Tested: guest_sce_net_tests round-trips fe80::1%2 (same address bytes
as fe80::1, scope ID 2) and ff02::1%4294967295, checks the exact
buffer size and ENOSPC one byte short, the unscoped forms, four
rejected texts with untouched outputs, EINVAL for null pointers, and
that an interface-name zone and AF_INET throw.
sceNetSysctl threw. Its restored argument list is FreeBSD's sysctl(3)
(name, name length, old value and size, new value and size), so it now
calls libkernel's sysctl. It answers exactly the MIBs libkernel models,
and a failure is returned as the library's other calls return one
(0x80410100 | errno, with the errno set).
Tested: guest_sce_net_tests reads hw.pagesize (6.7) through both
sysctl and sceNetSysctl and gets the same value and size, including a
size-only query. A one-level name gives EINVAL, a null name EFAULT, a
write EPERM and a one-byte buffer ENOMEM, each as 0x804101xx with the
errno set.
sceNetSocketInternal, sceNetGetsockoptEx and sceNetSetsockoptEx threw.
Their argument lists, restored with the rest of the firmware exports in
4d6942f1e, are the same as sceNetSocket, sceNetGetsockopt and
sceNetSetsockopt, so they now call those. Nothing in the tree records
how the console tells the entry points apart (for example by the
calling process); the debt file says so.
Tested: guest_sce_net_tests creates a datagram socket through
sceNetSocketInternal, sets SO_REUSEADDR and SO_RCVTIMEO through
sceNetSetsockoptEx and reads them back through sceNetGetsockoptEx and
sceNetGetsockopt with the same values. On a closed socket and on -1,
and for an unsupported family and type, each Ex/Internal call returns
the same error and errno as its public twin.
FreeBSD 11.4's soreceive_generic waits in sbwait (sys/kern/uipc_socket.c,
uipc_sockbuf.c), whose msleep_sbt on so_rcv.sb_timeo returns EWOULDBLOCK
(35) for every socket type when SO_RCVTIMEO expires. On Windows the host
socket is non-blocking and native_transfer enforces the timeout itself,
but it reported WSAETIMEDOUT, which only datagram receives turned into
EAGAIN; a stream receive failed with ETIMEDOUT (60). A receive that
reaches the stored timeout now reports WSAEWOULDBLOCK, so every
receive returns EAGAIN, and a host WSAETIMEDOUT still maps to ETIMEDOUT.
A PS5 (firmware 13.60) allocates with any flag value, zero-fills only for
flag 2 and returns 64-byte aligned blocks. Record the measured behaviour of
sceNetGetIfName, sceNetGetRandom, sceNetGetSystemTime, sceNetMemoryFree and
sceNetUsleep, which already match the console, and keep only their open gaps.
Co-Authored-By: Claude Opus 5.5 <[email protected]>