From c76562b6709fee5eff8a6a779be41c0bce661fd7 Mon Sep 17 00:00:00 2001 From: Mel Gorman Date: Tue, 31 Jul 2012 16:44:41 -0700 Subject: [PATCH] netvm: prevent a stream-specific deadlock This patch series is based on top of "Swap-over-NBD without deadlocking v15" as it depends on the same reservation of PF_MEMALLOC reserves logic. When a user or administrator requires swap for their application, they create a swap partition and file, format it with mkswap and activate it with swapon. In diskless systems this is not an option so if swap if required then swapping over the network is considered. The two likely scenarios are when blade servers are used as part of a cluster where the form factor or maintenance costs do not allow the use of disks and thin clients. The Linux Terminal Server Project recommends the use of the Network Block Device (NBD) for swap but this is not always an option. There is no guarantee that the network attached storage (NAS) device is running Linux or supports NBD. However, it is likely that it supports NFS so there are users that want support for swapping over NFS despite any performance concern. Some distributions currently carry patches that support swapping over NFS but it would be preferable to support it in the mainline kernel. Patch 1 avoids a stream-specific deadlock that potentially affects TCP. Patch 2 is a small modification to SELinux to avoid using PFMEMALLOC reserves. Patch 3 adds three helpers for filesystems to handle swap cache pages. For example, page_file_mapping() returns page->mapping for file-backed pages and the address_space of the underlying swap file for swap cache pages. Patch 4 adds two address_space_operations to allow a filesystem to pin all metadata relevant to a swapfile in memory. Upon successful activation, the swapfile is marked SWP_FILE and the address space operation ->direct_IO is used for writing and ->readpage for reading in swap pages. Patch 5 notes that patch 3 is bolting filesystem-specific-swapfile-support onto the side and that the default handlers have different information to what is available to the filesystem. This patch refactors the code so that there are generic handlers for each of the new address_space operations. Patch 6 adds an API to allow a vector of kernel addresses to be translated to struct pages and pinned for IO. Patch 7 adds support for using highmem pages for swap by kmapping the pages before calling the direct_IO handler. Patch 8 updates NFS to use the helpers from patch 3 where necessary. Patch 9 avoids setting PF_private on PG_swapcache pages within NFS. Patch 10 implements the new swapfile-related address_space operations for NFS and teaches the direct IO handler how to manage kernel addresses. Patch 11 prevents page allocator recursions in NFS by using GFP_NOIO where appropriate. Patch 12 fixes a NULL pointer dereference that occurs when using swap-over-NFS. With the patches applied, it is possible to mount a swapfile that is on an NFS filesystem. Swap performance is not great with a swap stress test taking roughly twice as long to complete than if the swap device was backed by NBD. This patch: netvm: prevent a stream-specific deadlock It could happen that all !SOCK_MEMALLOC sockets have buffered so much data that we're over the global rmem limit. This will prevent SOCK_MEMALLOC buffers from receiving data, which will prevent userspace from running, which is needed to reduce the buffered data. Fix this by exempting the SOCK_MEMALLOC sockets from the rmem limit. Once this change it applied, it is important that sockets that set SOCK_MEMALLOC do not clear the flag until the socket is being torn down. If this happens, a warning is generated and the tokens reclaimed to avoid accounting errors until the bug is fixed. [davem@davemloft.net: Warning about clearing SOCK_MEMALLOC] Signed-off-by: Peter Zijlstra Signed-off-by: Mel Gorman Acked-by: David S. Miller Acked-by: Rik van Riel Cc: Trond Myklebust Cc: Neil Brown Cc: Christoph Hellwig Cc: Mike Christie Cc: Eric B Munson Cc: Sebastian Andrzej Siewior Cc: Mel Gorman Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/net/sock.h | 8 +++++--- net/caif/caif_socket.c | 2 +- net/core/sock.c | 14 +++++++++++++- net/ipv4/tcp_input.c | 21 +++++++++++---------- net/sctp/ulpevent.c | 3 ++- 5 files changed, 32 insertions(+), 16 deletions(-) diff --git a/include/net/sock.h b/include/net/sock.h index 43a470d40d76..b3730239bf18 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1329,12 +1329,14 @@ static inline bool sk_wmem_schedule(struct sock *sk, int size) __sk_mem_schedule(sk, size, SK_MEM_SEND); } -static inline bool sk_rmem_schedule(struct sock *sk, int size) +static inline bool +sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, unsigned int size) { if (!sk_has_account(sk)) return true; - return size <= sk->sk_forward_alloc || - __sk_mem_schedule(sk, size, SK_MEM_RECV); + return size<= sk->sk_forward_alloc || + __sk_mem_schedule(sk, size, SK_MEM_RECV) || + skb_pfmemalloc(skb); } static inline void sk_mem_reclaim(struct sock *sk) diff --git a/net/caif/caif_socket.c b/net/caif/caif_socket.c index 78f1cdad5b33..095259f83902 100644 --- a/net/caif/caif_socket.c +++ b/net/caif/caif_socket.c @@ -141,7 +141,7 @@ static int caif_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) err = sk_filter(sk, skb); if (err) return err; - if (!sk_rmem_schedule(sk, skb->truesize) && rx_flow_is_on(cf_sk)) { + if (!sk_rmem_schedule(sk, skb, skb->truesize) && rx_flow_is_on(cf_sk)) { set_rx_flow_off(cf_sk); net_dbg_ratelimited("sending flow OFF due to rmem_schedule\n"); caif_flow_ctrl(sk, CAIF_MODEMCMD_FLOW_OFF_REQ); diff --git a/net/core/sock.c b/net/core/sock.c index 32fdcd2d6e8f..6b654b3ddfda 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -295,6 +295,18 @@ void sk_clear_memalloc(struct sock *sk) sock_reset_flag(sk, SOCK_MEMALLOC); sk->sk_allocation &= ~__GFP_MEMALLOC; static_key_slow_dec(&memalloc_socks); + + /* + * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward + * progress of swapping. However, if SOCK_MEMALLOC is cleared while + * it has rmem allocations there is a risk that the user of the + * socket cannot make forward progress due to exceeding the rmem + * limits. By rights, sk_clear_memalloc() should only be called + * on sockets being torn down but warn and reset the accounting if + * that assumption breaks. + */ + if (WARN_ON(sk->sk_forward_alloc)) + sk_mem_reclaim(sk); } EXPORT_SYMBOL_GPL(sk_clear_memalloc); @@ -396,7 +408,7 @@ int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) if (err) return err; - if (!sk_rmem_schedule(sk, skb->truesize)) { + if (!sk_rmem_schedule(sk, skb, skb->truesize)) { atomic_inc(&sk->sk_drops); return -ENOBUFS; } diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index a356e1fecf9a..00b91b4b8665 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -4351,19 +4351,20 @@ static void tcp_ofo_queue(struct sock *sk) static bool tcp_prune_ofo_queue(struct sock *sk); static int tcp_prune_queue(struct sock *sk); -static int tcp_try_rmem_schedule(struct sock *sk, unsigned int size) +static int tcp_try_rmem_schedule(struct sock *sk, struct sk_buff *skb, + unsigned int size) { if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || - !sk_rmem_schedule(sk, size)) { + !sk_rmem_schedule(sk, skb, size)) { if (tcp_prune_queue(sk) < 0) return -1; - if (!sk_rmem_schedule(sk, size)) { + if (!sk_rmem_schedule(sk, skb, size)) { if (!tcp_prune_ofo_queue(sk)) return -1; - if (!sk_rmem_schedule(sk, size)) + if (!sk_rmem_schedule(sk, skb, size)) return -1; } } @@ -4418,7 +4419,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb) TCP_ECN_check_ce(tp, skb); - if (unlikely(tcp_try_rmem_schedule(sk, skb->truesize))) { + if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) { NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPOFODROP); __kfree_skb(skb); return; @@ -4552,17 +4553,17 @@ static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size) { - struct sk_buff *skb; + struct sk_buff *skb = NULL; struct tcphdr *th; bool fragstolen; - if (tcp_try_rmem_schedule(sk, size + sizeof(*th))) - goto err; - skb = alloc_skb(size + sizeof(*th), sk->sk_allocation); if (!skb) goto err; + if (tcp_try_rmem_schedule(sk, skb, size + sizeof(*th))) + goto err_free; + th = (struct tcphdr *)skb_put(skb, sizeof(*th)); skb_reset_transport_header(skb); memset(th, 0, sizeof(*th)); @@ -4633,7 +4634,7 @@ static void tcp_data_queue(struct sock *sk, struct sk_buff *skb) if (eaten <= 0) { queue_and_out: if (eaten < 0 && - tcp_try_rmem_schedule(sk, skb->truesize)) + tcp_try_rmem_schedule(sk, skb, skb->truesize)) goto drop; eaten = tcp_queue_rcv(sk, skb, 0, &fragstolen); diff --git a/net/sctp/ulpevent.c b/net/sctp/ulpevent.c index 33d894776192..10c018a5b9fe 100644 --- a/net/sctp/ulpevent.c +++ b/net/sctp/ulpevent.c @@ -702,7 +702,8 @@ struct sctp_ulpevent *sctp_ulpevent_make_rcvmsg(struct sctp_association *asoc, if (rx_count >= asoc->base.sk->sk_rcvbuf) { if ((asoc->base.sk->sk_userlocks & SOCK_RCVBUF_LOCK) || - (!sk_rmem_schedule(asoc->base.sk, chunk->skb->truesize))) + (!sk_rmem_schedule(asoc->base.sk, chunk->skb, + chunk->skb->truesize))) goto fail; }