| created | 2024-02-07T04:55:16Z |
|---|---|
| begin | 2024-02-03T20:07:19Z |
| end | 2024-02-03T22:50:09Z |
| path | src/sys |
| commits | 2 |
| date | 2024-02-03T22:50:08Z | |||
|---|---|---|---|---|
| author | mvs | |||
| files | src/sys/kern/uipc_socket.c | log | diff | annotate |
| src/sys/kern/uipc_socket2.c | log | diff | annotate | |
| message |
Rework socket buffers locking for shared netlock. Shared netlock is not sufficient to call so{r,w}wakeup(). The following sowakeup() modifies `sb_flags' and knote(9) stuff. Unfortunately, we can't call so{r,w}wakeup() with `inp_mtx' mutex(9) because sowakeup() also calls pgsigio() which grabs kernel lock. However, `so*_filtops' callbacks only perform read-only access to the socket stuff, so it is enough to hold shared netlock only, but the klist stuff needs to be protected. This diff introduces `sb_mtx' mutex(9) to protect sockbuf. This time `sb_mtx' used to protect only `sb_flags' and `sb_klist'. Now we have soassertlocked_readonly() and soassertlocked(). The first one is happy if only shared netlock is held, meanwhile the second wants `so_lock' or pru_lock() be held together with shared netlock. To keep soassertlocked*() assertions soft, we need to know mutex(9) state, so new mtx_owned() macro was introduces. Also, the new optional (*pru_locked)() handler brings the state of pru_lock(). Tests and ok from bluhm. |
|||
| date | 2024-02-03T22:50:09Z | |||
|---|---|---|---|---|
| author | mvs | |||
| files | src/sys/kern/uipc_syscalls.c | log | diff | annotate |
| src/sys/miscfs/fifofs/fifo_vnops.c | log | diff | annotate | |
| src/sys/netinet/ip_divert.c | log | diff | annotate | |
| src/sys/netinet/ip_divert.h | log | diff | annotate | |
| src/sys/netinet/ip_var.h | log | diff | annotate | |
| src/sys/netinet/raw_ip.c | log | diff | annotate | |
| src/sys/netinet/udp_usrreq.c | log | diff | annotate | |
| src/sys/netinet/udp_var.h | log | diff | annotate | |
| src/sys/netinet6/ip6_divert.c | log | diff | annotate | |
| src/sys/netinet6/ip6_mroute.c | log | diff | annotate | |
| src/sys/netinet6/ip6_var.h | log | diff | annotate | |
| src/sys/netinet6/raw_ip6.c | log | diff | annotate | |
| src/sys/sys/mutex.h | log | diff | annotate | |
| src/sys/sys/protosw.h | log | diff | annotate | |
| src/sys/sys/socketvar.h | log | diff | annotate | |
| message |
Rework socket buffers locking for shared netlock. Shared netlock is not sufficient to call so{r,w}wakeup(). The following sowakeup() modifies `sb_flags' and knote(9) stuff. Unfortunately, we can't call so{r,w}wakeup() with `inp_mtx' mutex(9) because sowakeup() also calls pgsigio() which grabs kernel lock. However, `so*_filtops' callbacks only perform read-only access to the socket stuff, so it is enough to hold shared netlock only, but the klist stuff needs to be protected. This diff introduces `sb_mtx' mutex(9) to protect sockbuf. This time `sb_mtx' used to protect only `sb_flags' and `sb_klist'. Now we have soassertlocked_readonly() and soassertlocked(). The first one is happy if only shared netlock is held, meanwhile the second wants `so_lock' or pru_lock() be held together with shared netlock. To keep soassertlocked*() assertions soft, we need to know mutex(9) state, so new mtx_owned() macro was introduces. Also, the new optional (*pru_locked)() handler brings the state of pru_lock(). Tests and ok from bluhm. |
|||