Re: [PATCH] net: fix race in the receive/select

From: Eric Dumazet
Date: Thu Jun 25 2009 - 21:59:53 EST

Next message: Linus Torvalds: "Re: [PATCH v2] IA64 Compilation Error Fix for Intel IOMMU IdentityMapping Support"
Previous message: Chris Wright: "Re: [PATCH v2] IA64 Compilation Error Fix for Intel IOMMU IdentityMapping Support"
In reply to: Davide Libenzi: "Re: [PATCH] net: fix race in the receive/select"
Next in thread: Davide Libenzi: "Re: [PATCH] net: fix race in the receive/select"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Davide Libenzi a écrit :
> On Thu, 25 Jun 2009, Oleg Nesterov wrote:
>
>> Can't really comment this patch, except this all looks reasonable to me.
>> Add more CCs.
>
> While this can work, IMO it'd be cleaner to have the smp_mb() moved from
> fs/select.c to the ->poll() function.
> Having a barrier that matches another one in another susbsystem, because
> of the special locking logic of such subsystem, is not too shiny IMHO.
>

Yes but barrier is necessary only if add_wait_queue() was actually called, and __pollwait()
does this call.

Adding a plain smp_mb() in tcp_poll() for example would slowdown select()/poll() with NULL
timeout.

Adding a cond test before smp_mb() in tcp_poll() (and other ->poll() functions)
would be litle bit overkill too...

I believe this race was not existent in the past because spin_unlock() had a memory barrier,
and we changed this to a plain memory write at some point...

Most add_wait_queue() calls are followed by a call to set_current_state()
so a proper smp_mb() is explicitly included.

>
>
>
>> On 06/25, Jiri Olsa wrote:
>>> Adding memory barrier to the __pollwait function paired with
>>> receive callbacks. The smp_mb__after_lock define is added,
>>> since {read|write|spin}_lock() on x86 are full memory barriers.
>>>
>>> The race fires, when following code paths meet, and the tp->rcv_nxt and
>>> __add_wait_queue updates stay in CPU caches.
>>>
>>>
>>> CPU1 CPU2
>>>
>>> sys_select receive packet
>>> ... ...
>>> __add_wait_queue update tp->rcv_nxt
>>> ... ...
>>> tp->rcv_nxt check sock_def_readable
>>> ... {
>>> schedule ...
>>> if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> wake_up_interruptible(sk->sk_sleep)
>>> ...
>>> }
>>>
>>> If there was no cache the code would work ok, since the wait_queue and
>>> rcv_nxt are opposit to each other.
>>>
>>> Meaning that once tp->rcv_nxt is updated by CPU2, the CPU1 either already
>>> passed the tp->rcv_nxt check and sleeps, or will get the new value for
>>> tp->rcv_nxt and will return with new data mask.
>>> In both cases the process (CPU1) is being added to the wait queue, so the
>>> waitqueue_active (CPU2) call cannot miss and will wake up CPU1.
>>>
>>> The bad case is when the __add_wait_queue changes done by CPU1 stay in its
>>> cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 will then
>>> endup calling schedule and sleep forever if there are no more data on the
>>> socket.
>>>
>>> wbr,
>>> jirka
>>>
>>>
>>> Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
>>>
>>> ---
>>> arch/x86/include/asm/spinlock.h | 3 +++
>>> fs/select.c | 4 ++++
>>> include/linux/spinlock.h | 5 +++++
>>> include/net/sock.h | 18 ++++++++++++++++++
>>> net/atm/common.c | 4 ++--
>>> net/core/sock.c | 8 ++++----
>>> net/dccp/output.c | 2 +-
>>> net/iucv/af_iucv.c | 2 +-
>>> net/rxrpc/af_rxrpc.c | 2 +-
>>> net/unix/af_unix.c | 2 +-
>>> 10 files changed, 40 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/arch/x86/include/asm/spinlock.h b/arch/x86/include/asm/spinlock.h
>>> index b7e5db8..39ecc5f 100644
>>> --- a/arch/x86/include/asm/spinlock.h
>>> +++ b/arch/x86/include/asm/spinlock.h
>>> @@ -302,4 +302,7 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
>>> #define _raw_read_relax(lock) cpu_relax()
>>> #define _raw_write_relax(lock) cpu_relax()
>>>
>>> +/* The {read|write|spin}_lock() on x86 are full memory barriers. */
>>> +#define smp_mb__after_lock() do { } while (0)
>>> +
>>> #endif /* _ASM_X86_SPINLOCK_H */
>>> diff --git a/fs/select.c b/fs/select.c
>>> index d870237..c4bd5f0 100644
>>> --- a/fs/select.c
>>> +++ b/fs/select.c
>>> @@ -219,6 +219,10 @@ static void __pollwait(struct file *filp, wait_queue_head_t *wait_address,
>>> init_waitqueue_func_entry(&entry->wait, pollwake);
>>> entry->wait.private = pwq;
>>> add_wait_queue(wait_address, &entry->wait);
>>> +
>>> + /* This memory barrier is paired with the smp_mb__after_lock
>>> + * in the sk_has_sleeper. */
>>> + smp_mb();
>>> }
>>>
>>> int poll_schedule_timeout(struct poll_wqueues *pwq, int state,
>>> diff --git a/include/linux/spinlock.h b/include/linux/spinlock.h
>>> index 252b245..ae053bd 100644
>>> --- a/include/linux/spinlock.h
>>> +++ b/include/linux/spinlock.h
>>> @@ -132,6 +132,11 @@ do { \
>>> #endif /*__raw_spin_is_contended*/
>>> #endif
>>>
>>> +/* The lock does not imply full memory barrier. */
>>> +#ifndef smp_mb__after_lock
>>> +#define smp_mb__after_lock() smp_mb()
>>> +#endif
>>> +
>>> /**
>>> * spin_unlock_wait - wait until the spinlock gets unlocked
>>> * @lock: the spinlock in question.
>>> diff --git a/include/net/sock.h b/include/net/sock.h
>>> index 352f06b..7fbb143 100644
>>> --- a/include/net/sock.h
>>> +++ b/include/net/sock.h
>>> @@ -1241,6 +1241,24 @@ static inline int sk_has_allocations(const struct sock *sk)
>>> return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
>>> }
>>>
>>> +/**
>>> + * sk_has_sleeper - check if there are any waiting processes
>>> + * @sk: socket
>>> + *
>>> + * Returns true if socket has waiting processes
>>> + */
>>> +static inline int sk_has_sleeper(struct sock *sk)
>>> +{
>>> + /*
>>> + * We need to be sure we are in sync with the
>>> + * add_wait_queue modifications to the wait queue.
>>> + *
>>> + * This memory barrier is paired in the __pollwait.
>>> + */
>>> + smp_mb__after_lock();
>>> + return sk->sk_sleep && waitqueue_active(sk->sk_sleep);
>>> +}
>>> +
>>> /*
>>> * Queue a received datagram if it will fit. Stream and sequenced
>>> * protocols can't normally use this as they need to fit buffers in
>>> diff --git a/net/atm/common.c b/net/atm/common.c
>>> index c1c9793..67a8642 100644
>>> --- a/net/atm/common.c
>>> +++ b/net/atm/common.c
>>> @@ -92,7 +92,7 @@ static void vcc_sock_destruct(struct sock *sk)
>>> static void vcc_def_wakeup(struct sock *sk)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up(sk->sk_sleep);
>>> read_unlock(&sk->sk_callback_lock);
>>> }
>>> @@ -110,7 +110,7 @@ static void vcc_write_space(struct sock *sk)
>>> read_lock(&sk->sk_callback_lock);
>>>
>>> if (vcc_writable(sk)) {
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible(sk->sk_sleep);
>>>
>>> sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
>>> diff --git a/net/core/sock.c b/net/core/sock.c
>>> index b0ba569..6354863 100644
>>> --- a/net/core/sock.c
>>> +++ b/net/core/sock.c
>>> @@ -1715,7 +1715,7 @@ EXPORT_SYMBOL(sock_no_sendpage);
>>> static void sock_def_wakeup(struct sock *sk)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible_all(sk->sk_sleep);
>>> read_unlock(&sk->sk_callback_lock);
>>> }
>>> @@ -1723,7 +1723,7 @@ static void sock_def_wakeup(struct sock *sk)
>>> static void sock_def_error_report(struct sock *sk)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible_poll(sk->sk_sleep, POLLERR);
>>> sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
>>> read_unlock(&sk->sk_callback_lock);
>>> @@ -1732,7 +1732,7 @@ static void sock_def_error_report(struct sock *sk)
>>> static void sock_def_readable(struct sock *sk, int len)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible_sync_poll(sk->sk_sleep, POLLIN |
>>> POLLRDNORM | POLLRDBAND);
>>> sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
>>> @@ -1747,7 +1747,7 @@ static void sock_def_write_space(struct sock *sk)
>>> * progress. --DaveM
>>> */
>>> if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible_sync_poll(sk->sk_sleep, POLLOUT |
>>> POLLWRNORM | POLLWRBAND);
>>>
>>> diff --git a/net/dccp/output.c b/net/dccp/output.c
>>> index c0e88c1..c96119f 100644
>>> --- a/net/dccp/output.c
>>> +++ b/net/dccp/output.c
>>> @@ -196,7 +196,7 @@ void dccp_write_space(struct sock *sk)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>>
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible(sk->sk_sleep);
>>> /* Should agree with poll, otherwise some programs break */
>>> if (sock_writeable(sk))
>>> diff --git a/net/iucv/af_iucv.c b/net/iucv/af_iucv.c
>>> index 6be5f92..ba0149d 100644
>>> --- a/net/iucv/af_iucv.c
>>> +++ b/net/iucv/af_iucv.c
>>> @@ -306,7 +306,7 @@ static inline int iucv_below_msglim(struct sock *sk)
>>> static void iucv_sock_wake_msglim(struct sock *sk)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible_all(sk->sk_sleep);
>>> sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
>>> read_unlock(&sk->sk_callback_lock);
>>> diff --git a/net/rxrpc/af_rxrpc.c b/net/rxrpc/af_rxrpc.c
>>> index eac5e7b..60e0e38 100644
>>> --- a/net/rxrpc/af_rxrpc.c
>>> +++ b/net/rxrpc/af_rxrpc.c
>>> @@ -63,7 +63,7 @@ static void rxrpc_write_space(struct sock *sk)
>>> _enter("%p", sk);
>>> read_lock(&sk->sk_callback_lock);
>>> if (rxrpc_writable(sk)) {
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible(sk->sk_sleep);
>>> sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
>>> }
>>> diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
>>> index 36d4e44..143143a 100644
>>> --- a/net/unix/af_unix.c
>>> +++ b/net/unix/af_unix.c
>>> @@ -315,7 +315,7 @@ static void unix_write_space(struct sock *sk)
>>> {
>>> read_lock(&sk->sk_callback_lock);
>>> if (unix_writable(sk)) {
>>> - if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
>>> + if (sk_has_sleeper(sk))
>>> wake_up_interruptible_sync(sk->sk_sleep);
>>> sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
>>> }
>> --
>> To unsubscribe from this list: send the line "unsubscribe netdev" in
>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
>
> - Davide
>
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Linus Torvalds: "Re: [PATCH v2] IA64 Compilation Error Fix for Intel IOMMU IdentityMapping Support"
Previous message: Chris Wright: "Re: [PATCH v2] IA64 Compilation Error Fix for Intel IOMMU IdentityMapping Support"
In reply to: Davide Libenzi: "Re: [PATCH] net: fix race in the receive/select"
Next in thread: Davide Libenzi: "Re: [PATCH] net: fix race in the receive/select"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]