[PATCH v2 0/1] MR6021: nsiproxy.sys: Use net.inet.tcp.pcblist64 to enumerate connections on macOS.

[Tim Clem (＠tclem)]

The handles returned by libproc (namely struct socket_info's soi_pcb)
use all 64 bits, but the ones from the pcblist sysctl are truncated
to 32. That makes find_owning_pid fail. The pcblist64 sysctl was
added in macOS 10.6 and returns handles that match those from
libproc.

--

There does not seem to be a MIB constant for pcblist64, so I had to fetch it with sysctlbyname.

--
  v2: nsiproxy.sys: Use net.inet.tcp.pcblist64 to enumerate connections on macOS.

https://gitlab.winehq.org/wine/wine/-/merge_requests/6021

[PATCH v2 1/1] nsiproxy.sys: Use net.inet.tcp.pcblist64 to enumerate connections on macOS.

[Tim Clem]

From: Tim Clem

The handles returned by libproc (namely struct socket_info's soi_pcb)
use all 64 bits, but the ones from the pcblist sysctl are truncated
to 32. That makes find_owning_pid fail. The pcblist64 sysctl was
added in macOS 10.6 and returns handles that match those from
libproc.
---
 dlls/nsiproxy.sys/tcp.c | 24 ++++++++++++++++++++----
 1 file changed, 20 insertions(+), 4 deletions(-)

diff --git a/dlls/nsiproxy.sys/tcp.c b/dlls/nsiproxy.sys/tcp.c
index 5734c4d9ee0..f9aa516e9f9 100644
--- a/dlls/nsiproxy.sys/tcp.c
+++ b/dlls/nsiproxy.sys/tcp.c
@@ -625,12 +625,24 @@ static NTSTATUS tcp_conns_enumerate_all( UINT filter, struct nsi_tcp_conn_key *k
     }
 #elif defined(HAVE_SYS_SYSCTL_H) && defined(TCPCTL_PCBLIST) && defined(HAVE_STRUCT_XINPGEN)
     {
-        int mib[] = { CTL_NET, PF_INET, IPPROTO_TCP, TCPCTL_PCBLIST };
         size_t len = 0;
         char *buf = NULL;
         struct xinpgen *xig, *orig_xig;

-        if (sysctl( mib, ARRAY_SIZE(mib), NULL, &len, NULL, 0 ) < 0)
+#ifdef __APPLE__
+        static int mib[CTL_MAXNAME];
+        static size_t mib_len = 0;
+        if (mib_len == 0)
+        {
+            mib_len = CTL_MAXNAME;
+            sysctlnametomib( "net.inet.tcp.pcblist64", mib, &mib_len );
+        }
+#else
+        int mib[] = { CTL_NET, PF_INET, IPPROTO_TCP, TCPCTL_PCBLIST };
+        size_t mib_len = ARRAY_SIZE(mib);
+#endif
+
+        if (sysctl( mib, mib_len, NULL, &len, NULL, 0 ) < 0)
         {
             ERR( "Failure to read net.inet.tcp.pcblist via sysctl\n" );
             status = STATUS_NOT_SUPPORTED;
@@ -644,7 +656,7 @@ static NTSTATUS tcp_conns_enumerate_all( UINT filter, struct nsi_tcp_conn_key *k
             goto err;
         }

-        if (sysctl( mib, ARRAY_SIZE(mib), buf, &len, NULL, 0 ) < 0)
+        if (sysctl( mib, mib_len, buf, &len, NULL, 0 ) < 0)
         {
             ERR( "Failure to read net.inet.tcp.pcblist via sysctl\n" );
             status = STATUS_NOT_SUPPORTED;
@@ -664,7 +676,11 @@ static NTSTATUS tcp_conns_enumerate_all( UINT filter, struct nsi_tcp_conn_key *k
              xig->xig_len > sizeof(struct xinpgen);
              xig = (struct xinpgen *)((char *)xig + xig->xig_len))
         {
-#if __FreeBSD_version >= 1200026
+#ifdef __APPLE__
+            struct xtcpcb64 *tcp = (struct xtcpcb64 *)xig;
+            struct xinpcb64 *in = &tcp->xt_inpcb;
+            struct xsocket64 *sock = &in->xi_socket;
+#elif __FreeBSD_version >= 1200026
             struct xtcpcb *tcp = (struct xtcpcb *)xig;
             struct xinpcb *in = &tcp->xt_inp;
             struct xsocket *sock = &in->xi_socket;

[PATCH v2 1/1] nsiproxy.sys: Use net.inet.tcp.pcblist64 to enumerate connections on macOS.

[Marvin]

Hi,

It looks like your patch introduced the new failures shown below.
Please investigate and fix them before resubmitting your patch.
If they are not new, fixing them anyway would help a lot. Otherwise
please ask for the known failures list to be updated.

The tests also ran into some preexisting test failures. If you know how
to fix them that would be helpful. See the TestBot job for the details:

The full results can be found at:
https://testbot.winehq.org/JobDetails.pl?Key=146936

Your paranoid android.

=== debian11b (64 bit WoW report) ===

ddraw:
ddraw7.c:3734: Test failed: Expected message 0x5, but didn't receive it.
ddraw7.c:3736: Test failed: Expected screen size 1024x768, got 0x0.
ddraw7.c:3743: Test failed: Expected (0,0)-(1024,768), got (-32000,-32000)-(-31840,-31969).
ddraw7.c:3813: Test failed: Expected (0,0)-(640,480), got (-32000,-32000)-(-31840,-31969).
ddraw7.c:3838: Test failed: Expected (0,0)-(640,480), got (-32000,-32000)-(-31840,-31969).

[Tim Clem (＠tclem)]

On Tue Jul  9 06:42:04 2024 +0000, Huw Davies wrote:
> > Looks fine.
>
> Actually, no it doesn't.  The first call returns the length, you'd
> presumably need to replace the second call further down (perhaps even
> more reason to cache the mib).

Whoa, whoops. Thanks, missed that when I was cleaning up the patch. I'm now caching the mib, which cleans things up a bit as well.