Select Git revision
Forked from
hardware-enablement / Rockchip upstream enablement efforts / linux
Source project has a limited visibility.
-
Eric Dumazet authoredsk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by:
Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>Eric Dumazet authoredsk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by:
Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
af_unix.c 52.47 KiB
/*
* NET4: Implementation of BSD Unix domain sockets.
*
* Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk>
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*
* Fixes:
* Linus Torvalds : Assorted bug cures.
* Niibe Yutaka : async I/O support.
* Carsten Paeth : PF_UNIX check, address fixes.
* Alan Cox : Limit size of allocated blocks.
* Alan Cox : Fixed the stupid socketpair bug.
* Alan Cox : BSD compatibility fine tuning.
* Alan Cox : Fixed a bug in connect when interrupted.
* Alan Cox : Sorted out a proper draft version of
* file descriptor passing hacked up from
* Mike Shaver's work.
* Marty Leisner : Fixes to fd passing
* Nick Nevin : recvmsg bugfix.
* Alan Cox : Started proper garbage collector
* Heiko EiBfeldt : Missing verify_area check
* Alan Cox : Started POSIXisms
* Andreas Schwab : Replace inode by dentry for proper
* reference counting
* Kirk Petersen : Made this a module
* Christoph Rohland : Elegant non-blocking accept/connect algorithm.
* Lots of bug fixes.
* Alexey Kuznetosv : Repaired (I hope) bugs introduces
* by above two patches.
* Andrea Arcangeli : If possible we block in connect(2)
* if the max backlog of the listen socket
* is been reached. This won't break
* old apps and it will avoid huge amount
* of socks hashed (this for unix_gc()
* performances reasons).
* Security fix that limits the max
* number of socks to 2*max_files and
* the number of skb queueable in the
* dgram receiver.
* Artur Skawina : Hash function optimizations
* Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8)
* Malcolm Beattie : Set peercred for socketpair
* Michal Ostrowski : Module initialization cleanup.
* Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT,
* the core infrastructure is doing that
* for all net proto families now (2.5.69+)
*
*
* Known differences from reference BSD that was tested:
*
* [TO FIX]
* ECONNREFUSED is not returned from one end of a connected() socket to the
* other the moment one end closes.
* fstat() doesn't return st_dev=0, and give the blksize as high water mark
* and a fake inode identifier (nor the BSD first socket fstat twice bug).
* [NOT TO FIX]
* accept() returns a path name even if the connecting socket has closed
* in the meantime (BSD loses the path and gives up).
* accept() returns 0 length path for an unbound connector. BSD returns 16
* and a null first byte in the path (but not for gethost/peername - BSD bug ??)
* socketpair(...SOCK_RAW..) doesn't panic the kernel.
* BSD af_unix apparently has connect forgetting to block properly.
* (need to check this with the POSIX spec in detail)
*
* Differences from 2.0.0-11-... (ANK)
* Bug fixes and improvements.