The unix socket code is using the result of csum_partial to
hash into a lookup table:

	unix_hash_fold(csum_partial(sunaddr, len, 0));

csum_partial is only guaranteed to produce something that can be
folded into a checksum, as its prototype explains:

 * returns a 32-bit number suitable for feeding into itself
 * or csum_tcpudp_magic

The 32bit value should not be used directly.

Depending on the alignment, the ppc64 csum_partial will return
different 32bit partial checksums that will fold into the same
16bit checksum.

This difference causes the following testcase (courtesy of
Gustavo) to sometimes fail:

#include <sys/socket.h>
#include <stdio.h>

int main()
{
	int fd = socket(PF_LOCAL, SOCK_STREAM|SOCK_CLOEXEC, 0);

	int i = 1;
	setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &i, 4);

	struct sockaddr addr;
	addr.sa_family = AF_LOCAL;
	bind(fd, &addr, 2);

	listen(fd, 128);

	struct sockaddr_storage ss;
	socklen_t sslen = (socklen_t)sizeof(ss);
	getsockname(fd, (struct sockaddr*)&ss, &sslen);

	fd = socket(PF_LOCAL, SOCK_STREAM|SOCK_CLOEXEC, 0);

	if (connect(fd, (struct sockaddr*)&ss, sslen) == -1){
		perror(NULL);
		return 1;
	}
	printf("OK\n");
	return 0;
}

As suggested by davem, fix this by using csum_fold to fold the
partial 32bit checksum into a 16bit checksum before using it.
Signed-off-by: NAnton Blanchard <anton@samba.org>
Cc: stable@vger.kernel.org
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The original documentation was very unclear.

The code fix is presumably related to the formerly unclear
documentation: SOCK_TIMESTAMPING_RX_SOFTWARE has no effect on
__sock_recv_timestamp's behavior, so calling __sock_recv_ts_and_drops
from sock_recv_ts_and_drops if only SOCK_TIMESTAMPING_RX_SOFTWARE is
set is pointless.  This should have no user-observable effect.
Signed-off-by: NAndy Lutomirski <luto@amacapital.net>
Acked-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

With -Werror=array-bounds, gcc v4.7.x warns that in phy_find_valid(), the
settings[] "array subscript is above array bounds", I think because idx is
a signed integer and if the caller supplied idx < 0, we pass the guard but
still reference out of bounds.

Fix this by making idx unsigned here and elsewhere.
Signed-off-by: NBjorn Helgaas <bhelgaas@google.com>
Acked-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Quoting Alexander Aring:
  While fragmentation and unloading of 6lowpan module I got this kernel Oops
  after few seconds:

  BUG: unable to handle kernel paging request at f88bbc30
  [..]
  Modules linked in: ipv6 [last unloaded: 6lowpan]
  Call Trace:
   [<c012af4c>] ? call_timer_fn+0x54/0xb3
   [<c012aef8>] ? process_timeout+0xa/0xa
   [<c012b66b>] run_timer_softirq+0x140/0x15f

Problem is that incomplete frags are still around after unload; when
their frag expire timer fires, we get crash.

When a netns is removed (also done when unloading module), inet_frag
calls the evictor with 'force' argument to purge remaining frags.

The evictor loop terminates when accounted memory ('work') drops to 0
or the lru-list becomes empty.  However, the mem accounting is done
via percpu counters and may not be accurate, i.e. loop may terminate
prematurely.

Alter evictor to only stop once the lru list is empty when force is
requested.
Reported-by: NPhoebe Buckheister <phoebe.buckheister@itwm.fraunhofer.de>
Reported-by: NAlexander Aring <alex.aring@gmail.com>
Tested-by: NAlexander Aring <alex.aring@gmail.com>
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Eric Hugne says:

====================
tipc: refcount and memory leak fixes

v3: Remove error logging from data path completely. Rebased on top of
    latest net merge.

v2: Drop specific -ENOMEM logging in patch #1 (tipc: allow connection
    shutdown callback to be invoked in advance) And add a general error
    message if an internal server tries to send a message on a
    closed/nonexisting connection.

In addition to the fix for refcount leak and memory leak during
module removal, we also fix a problem where the topology server
listening socket where unexpectedly closed. We also eliminate an
unnecessary context switch during accept()/recvmsg() for nonblocking
sockets.

It might be good to include this patchset in stable aswell. After the
v3 rebase on latest merge from net all patches apply cleanly on that
tree.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Failure to schedule a TIPC tasklet with tipc_k_signal because the
tasklet handler is disabled is not an error. It means TIPC is
currently in the process of shutting down. We remove the error
logging in this case.
Signed-off-by: NErik Hugne <erik.hugne@ericsson.com>
Reviewed-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the TIPC module is removed, the tasklet handler is disabled
before all other subsystems. This will cause lingering publications
in the name table because the node_down tasklets responsible to
clean up publications from an unreachable node will never run.
When the name table is shut down, these publications are detected
and an error message is logged:
tipc: nametbl_stop(): orphaned hash chain detected
This is actually a memory leak, introduced with commit
993b858e ("tipc: correct the order
of stopping services at rmmod")

Instead of just logging an error and leaking memory, we free
the orphaned entries during nametable shutdown.
Signed-off-by: NErik Hugne <erik.hugne@ericsson.com>
Reviewed-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a topology server subscriber is disconnected, the associated
connection id is set to zero. A check vs zero is then done in the
subscription timeout function to see if the subscriber have been
shut down. This is unnecessary, because all subscription timers
will be cancelled when a subscriber terminates. Setting the
connection id to zero is actually harmful because id zero is the
identity of the topology server listening socket, and can cause a
race that leads to this socket being closed instead.
Signed-off-by: NErik Hugne <erik.hugne@ericsson.com>
Acked-by: NYing Xue <ying.xue@windriver.com>
Reviewed-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When messages are received via tipc socket under non-block mode,
schedule_timeout() is called in tipc_wait_for_rcvmsg(), that is,
the process of receiving messages will be scheduled once although
timeout value passed to schedule_timeout() is 0. The same issue
exists in accept()/wait_for_accept(). To avoid this unnecessary
process switch, we only call schedule_timeout() if the timeout
value is non-zero.
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Reviewed-by: NErik Hugne <erik.hugne@ericsson.com>
Reviewed-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When tipc_conn_sendmsg() calls tipc_conn_lookup() to query a
connection instance, its reference count value is increased if
it's found. But subsequently if it's found that the connection is
closed, the work of sending message is not queued into its server
send workqueue, and the connection reference count is not decreased.
This will cause a reference count leak. To reproduce this problem,
an application would need to open and closes topology server
connections with high intensity.

We fix this by immediately decrementing the connection reference
count if a send fails due to the connection being closed.
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Acked-by: NErik Hugne <erik.hugne@ericsson.com>
Reviewed-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently connection shutdown callback function is called when
connection instance is released in tipc_conn_kref_release(), and
receiving packets and sending packets are running in different
threads. Even if connection is closed by the thread of receiving
packets, its shutdown callback may not be called immediately as
the connection reference count is non-zero at that moment. So,
although the connection is shut down by the thread of receiving
packets, the thread of sending packets doesn't know it. Before
its shutdown callback is invoked to tell the sending thread its
connection has been closed, the sending thread may deliver
messages by tipc_conn_sendmsg(), this is why the following error
information appears:

"Sending subscription event failed, no memory"

To eliminate it, allow connection shutdown callback function to
be called before connection id is removed in tipc_close_conn(),
which makes the sending thread know the truth in time that its
socket is closed so that it doesn't send message to it. We also
remove the "Sending XXX failed..." error reporting for topology
and config services.
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NErik Hugne <erik.hugne@ericsson.com>
Reviewed-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As pppol2tp_recv() never queues up packets to plain L2TP sockets,
pppol2tp_recvmsg() never returns data to userspace, thus making
the recv*() system calls unusable.

Instead of dropping packets when the L2TP socket isn't bound to a PPP
channel, this patch adds them to its reception queue.
Signed-off-by: NGuillaume Nault <g.nault@alphalink.fr>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit e0d4435f "l2tp: Update PPP-over-L2TP driver to work over L2TPv3"
broke the PPPOL2TP_SO_SENDSEQ setsockopt. The L2TP header length was
previously computed by pppol2tp_l2t_header_len() before each call to
l2tp_xmit_skb(). Now that header length is retrieved from the hdr_len
session field, this field must be updated every time the L2TP header
format is modified, or l2tp_xmit_skb() won't push the right amount of
data for the L2TP header.

This patch uses l2tp_session_set_header_len() to adjust hdr_len every
time sequencing is (de)activated from userspace (either by the
PPPOL2TP_SO_SENDSEQ setsockopt or the L2TP_ATTR_SEND_SEQ netlink
attribute).
Signed-off-by: NGuillaume Nault <g.nault@alphalink.fr>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It moves the state setting for query into rndis_filter_receive_response().
All callbacks including query-complete and status-callback are synchronized
by channel->inbound_lock. This prevents pentential race between them.
Signed-off-by: NHaiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When allocating RX buffers a fixed size is used, while freeing is based
on actually received bytes, resulting in the following kernel warning
when CONFIG_DMA_API_DEBUG is enabled:
 WARNING: CPU: 0 PID: 0 at lib/dma-debug.c:1051 check_unmap+0x258/0x894()
 macb e000b000.ethernet: DMA-API: device driver frees DMA memory with different size [device address=0x000000002d170040] [map size=1536 bytes] [unmap size=60 bytes]
 Modules linked in:
 CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.14.0-rc3-xilinx-00220-g49f84081ce4f #65
 [<c001516c>] (unwind_backtrace) from [<c0011df8>] (show_stack+0x10/0x14)
 [<c0011df8>] (show_stack) from [<c03c775c>] (dump_stack+0x7c/0xc8)
 [<c03c775c>] (dump_stack) from [<c00245cc>] (warn_slowpath_common+0x60/0x84)
 [<c00245cc>] (warn_slowpath_common) from [<c0024670>] (warn_slowpath_fmt+0x2c/0x3c)
 [<c0024670>] (warn_slowpath_fmt) from [<c0227d44>] (check_unmap+0x258/0x894)
 [<c0227d44>] (check_unmap) from [<c0228588>] (debug_dma_unmap_page+0x64/0x70)
 [<c0228588>] (debug_dma_unmap_page) from [<c02ab78c>] (gem_rx+0x118/0x170)
 [<c02ab78c>] (gem_rx) from [<c02ac4d4>] (macb_poll+0x24/0x94)
 [<c02ac4d4>] (macb_poll) from [<c031222c>] (net_rx_action+0x6c/0x188)
 [<c031222c>] (net_rx_action) from [<c0028a28>] (__do_softirq+0x108/0x280)
 [<c0028a28>] (__do_softirq) from [<c0028e8c>] (irq_exit+0x84/0xf8)
 [<c0028e8c>] (irq_exit) from [<c000f360>] (handle_IRQ+0x68/0x8c)
 [<c000f360>] (handle_IRQ) from [<c0008528>] (gic_handle_irq+0x3c/0x60)
 [<c0008528>] (gic_handle_irq) from [<c0012904>] (__irq_svc+0x44/0x78)
 Exception stack(0xc056df20 to 0xc056df68)
 df20: 00000001 c0577430 00000000 c0577430 04ce8e0d 00000002 edfce238 00000000
 df40: 04e20f78 00000002 c05981f4 00000000 00000008 c056df68 c0064008 c02d7658
 df60: 20000013 ffffffff
 [<c0012904>] (__irq_svc) from [<c02d7658>] (cpuidle_enter_state+0x54/0xf8)
 [<c02d7658>] (cpuidle_enter_state) from [<c02d77dc>] (cpuidle_idle_call+0xe0/0x138)
 [<c02d77dc>] (cpuidle_idle_call) from [<c000f660>] (arch_cpu_idle+0x8/0x3c)
 [<c000f660>] (arch_cpu_idle) from [<c006bec4>] (cpu_startup_entry+0xbc/0x124)
 [<c006bec4>] (cpu_startup_entry) from [<c053daec>] (start_kernel+0x350/0x3b0)
 ---[ end trace d5fdc38641bd3a11 ]---
 Mapped at:
  [<c0227184>] debug_dma_map_page+0x48/0x11c
  [<c02ab32c>] gem_rx_refill+0x154/0x1f8
  [<c02ac7b4>] macb_open+0x270/0x3e0
  [<c03152e0>] __dev_open+0x7c/0xfc
  [<c031554c>] __dev_change_flags+0x8c/0x140

Fixing this by passing the same size which is passed during mapping the
memory to the unmap function as well.
Signed-off-by: NSoren Brinkmann <soren.brinkmann@xilinx.com>
Reviewed-by: NBen Hutchings <ben@decadent.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

With CONFIG_DMA_API_DEBUG enabled the following warning is printed:
 WARNING: CPU: 0 PID: 619 at lib/dma-debug.c:1101 check_unmap+0x758/0x894()
 macb e000b000.ethernet: DMA-API: device driver failed to check map error[device address=0x000000002d171c02] [size=322 bytes] [mapped as single]
 Modules linked in:
 CPU: 0 PID: 619 Comm: udhcpc Not tainted 3.14.0-rc3-xilinx-00219-gd158fc7f #63
 [<c001516c>] (unwind_backtrace) from [<c0011df8>] (show_stack+0x10/0x14)
 [<c0011df8>] (show_stack) from [<c03c7714>] (dump_stack+0x7c/0xc8)
 [<c03c7714>] (dump_stack) from [<c00245cc>] (warn_slowpath_common+0x60/0x84)
 [<c00245cc>] (warn_slowpath_common) from [<c0024670>] (warn_slowpath_fmt+0x2c/0x3c)
 [<c0024670>] (warn_slowpath_fmt) from [<c0228244>] (check_unmap+0x758/0x894)
 [<c0228244>] (check_unmap) from [<c0228588>] (debug_dma_unmap_page+0x64/0x70)
 [<c0228588>] (debug_dma_unmap_page) from [<c02aba64>] (macb_interrupt+0x1f8/0x2dc)
 [<c02aba64>] (macb_interrupt) from [<c006c6e4>] (handle_irq_event_percpu+0x2c/0x178)
 [<c006c6e4>] (handle_irq_event_percpu) from [<c006c86c>] (handle_irq_event+0x3c/0x5c)
 [<c006c86c>] (handle_irq_event) from [<c006f548>] (handle_fasteoi_irq+0xb8/0x100)
 [<c006f548>] (handle_fasteoi_irq) from [<c006c148>] (generic_handle_irq+0x20/0x30)
 [<c006c148>] (generic_handle_irq) from [<c000f35c>] (handle_IRQ+0x64/0x8c)
 [<c000f35c>] (handle_IRQ) from [<c0008528>] (gic_handle_irq+0x3c/0x60)
 [<c0008528>] (gic_handle_irq) from [<c0012904>] (__irq_svc+0x44/0x78)
 Exception stack(0xed197f60 to 0xed197fa8)
 7f60: 00000134 60000013 bd94362e bd94362e be96b37c 00000014 fffffd72 00000122
 7f80: c000ebe4 ed196000 00000000 00000011 c032c0d8 ed197fa8 c0064008 c000ea20
 7fa0: 60000013 ffffffff
 [<c0012904>] (__irq_svc) from [<c000ea20>] (ret_fast_syscall+0x0/0x48)
 ---[ end trace 478f921d0d542d1e ]---
 Mapped at:
  [<c0227184>] debug_dma_map_page+0x48/0x11c
  [<c02aaca0>] macb_start_xmit+0x184/0x2a8
  [<c03143c0>] dev_hard_start_xmit+0x334/0x470
  [<c032c09c>] sch_direct_xmit+0x78/0x2f8
  [<c0314814>] __dev_queue_xmit+0x318/0x708

due to missing checks of the dma mapping. Add the appropriate checks to fix
this.
Signed-off-by: NSoren Brinkmann <soren.brinkmann@xilinx.com>
Reviewed-by: NBen Hutchings <ben@decadent.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While working on ec0223ec ("net: sctp: fix sctp_sf_do_5_1D_ce to
verify if we/peer is AUTH capable"), we noticed that there's a skb
memory leakage in the error path.

Running the same reproducer as in ec0223ec and by unconditionally
jumping to the error label (to simulate an error condition) in
sctp_sf_do_5_1D_ce() receive path lets kmemleak detector bark about
the unfreed chunk->auth_chunk skb clone:

Unreferenced object 0xffff8800b8f3a000 (size 256):
  comm "softirq", pid 0, jiffies 4294769856 (age 110.757s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    89 ab 75 5e d4 01 58 13 00 00 00 00 00 00 00 00  ..u^..X.........
  backtrace:
    [<ffffffff816660be>] kmemleak_alloc+0x4e/0xb0
    [<ffffffff8119f328>] kmem_cache_alloc+0xc8/0x210
    [<ffffffff81566929>] skb_clone+0x49/0xb0
    [<ffffffffa0467459>] sctp_endpoint_bh_rcv+0x1d9/0x230 [sctp]
    [<ffffffffa046fdbc>] sctp_inq_push+0x4c/0x70 [sctp]
    [<ffffffffa047e8de>] sctp_rcv+0x82e/0x9a0 [sctp]
    [<ffffffff815abd38>] ip_local_deliver_finish+0xa8/0x210
    [<ffffffff815a64af>] nf_reinject+0xbf/0x180
    [<ffffffffa04b4762>] nfqnl_recv_verdict+0x1d2/0x2b0 [nfnetlink_queue]
    [<ffffffffa04aa40b>] nfnetlink_rcv_msg+0x14b/0x250 [nfnetlink]
    [<ffffffff815a3269>] netlink_rcv_skb+0xa9/0xc0
    [<ffffffffa04aa7cf>] nfnetlink_rcv+0x23f/0x408 [nfnetlink]
    [<ffffffff815a2bd8>] netlink_unicast+0x168/0x250
    [<ffffffff815a2fa1>] netlink_sendmsg+0x2e1/0x3f0
    [<ffffffff8155cc6b>] sock_sendmsg+0x8b/0xc0
    [<ffffffff8155d449>] ___sys_sendmsg+0x369/0x380

What happens is that commit bbd0d598 clones the skb containing
the AUTH chunk in sctp_endpoint_bh_rcv() when having the edge case
that an endpoint requires COOKIE-ECHO chunks to be authenticated:

  ---------- INIT[RANDOM; CHUNKS; HMAC-ALGO] ---------->
  <------- INIT-ACK[RANDOM; CHUNKS; HMAC-ALGO] ---------
  ------------------ AUTH; COOKIE-ECHO ---------------->
  <-------------------- COOKIE-ACK ---------------------

When we enter sctp_sf_do_5_1D_ce() and before we actually get to
the point where we process (and subsequently free) a non-NULL
chunk->auth_chunk, we could hit the "goto nomem_init" path from
an error condition and thus leave the cloned skb around w/o
freeing it.

The fix is to centrally free such clones in sctp_chunk_destroy()
handler that is invoked from sctp_chunk_free() after all refs have
dropped; and also move both kfree_skb(chunk->auth_chunk) there,
so that chunk->auth_chunk is either NULL (since sctp_chunkify()
allocs new chunks through kmem_cache_zalloc()) or non-NULL with
a valid skb pointer. chunk->skb and chunk->auth_chunk are the
only skbs in the sctp_chunk structure that need to be handeled.

While at it, we should use consume_skb() for both. It is the same
as dev_kfree_skb() but more appropriately named as we are not
a device but a protocol. Also, this effectively replaces the
kfree_skb() from both invocations into consume_skb(). Functions
are the same only that kfree_skb() assumes that the frame was
being dropped after a failure (e.g. for tools like drop monitor),
usage of consume_skb() seems more appropriate in function
sctp_chunk_destroy() though.

Fixes: bbd0d598 ("[SCTP]: Implement the receive and verification of AUTH chunk")
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Cc: Vlad Yasevich <yasevich@gmail.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Acked-by: NVlad Yasevich <vyasevich@gmail.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are known issues for switching the drivers between ECM mode and
vendor mode. The interrup transfer may become abnormal. The hardware
may have the opportunity to die if you change the configuration without
unloading the current driver first, because all the control transfers
of the current driver would fail after the command of switching the
configuration.

Although to use the ecm driver and vendor driver independently is fine,
it may have problems to change the driver from one to the other by
switching the configuration. Additionally, now the vendor mode driver
is more powerful than the ECM driver. Thus, disable the ECM mode driver,
and let r8152 to set the configuration to vendor mode and reset the
device automatically.
Signed-off-by: NHayes Wang <hayeswang@realtek.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In kexec scenario, we failed to load the mlx4 driver in the
second kernel because the ownership bit was hold by the first
kernel without release correctly.

The patch adds shutdown() interface so that the ownership can
be released correctly in the first kernel. It also helps avoiding
EEH error happened during boot stage of the second kernel because
of undesired traffic, which can't be handled by hardware during
that stage on Power platform.
Signed-off-by: NGavin Shan <shangw@linux.vnet.ibm.com>
Tested-by: NWei Yang <weiyang@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

MLD queries are supposed to have an IPv6 link-local source address
according to RFC2710, section 4 and RFC3810, section 5.1.14. This patch
adds a sanity check to ignore such broken MLD queries.

Without this check, such malformed MLD queries can result in a
denial of service: The queries are ignored by any MLD listener
therefore they will not respond with an MLD report. However,
without this patch these malformed MLD queries would enable the
snooping part in the bridge code, potentially shutting down the
according ports towards these hosts for multicast traffic as the
bridge did not learn about these listeners.
Reported-by: NJan Stancek <jstancek@redhat.com>
Signed-off-by: NLinus Lüssing <linus.luessing@web.de>
Reviewed-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

I stumbled upon this very serious bug while hunting for another one,
it's a very subtle race condition between inet_frag_evictor,
inet_frag_intern and the IPv4/6 frag_queue and expire functions
(basically the users of inet_frag_kill/inet_frag_put).

What happens is that after a fragment has been added to the hash chain
but before it's been added to the lru_list (inet_frag_lru_add) in
inet_frag_intern, it may get deleted (either by an expired timer if
the system load is high or the timer sufficiently low, or by the
fraq_queue function for different reasons) before it's added to the
lru_list, then after it gets added it's a matter of time for the
evictor to get to a piece of memory which has been freed leading to a
number of different bugs depending on what's left there.

I've been able to trigger this on both IPv4 and IPv6 (which is normal
as the frag code is the same), but it's been much more difficult to
trigger on IPv4 due to the protocol differences about how fragments
are treated.

The setup I used to reproduce this is: 2 machines with 4 x 10G bonded
in a RR bond, so the same flow can be seen on multiple cards at the
same time. Then I used multiple instances of ping/ping6 to generate
fragmented packets and flood the machines with them while running
other processes to load the attacked machine.

*It is very important to have the _same flow_ coming in on multiple CPUs
concurrently. Usually the attacked machine would die in less than 30
minutes, if configured properly to have many evictor calls and timeouts
it could happen in 10 minutes or so.

An important point to make is that any caller (frag_queue or timer) of
inet_frag_kill will remove both the timer refcount and the
original/guarding refcount thus removing everything that's keeping the
frag from being freed at the next inet_frag_put.  All of this could
happen before the frag was ever added to the LRU list, then it gets
added and the evictor uses a freed fragment.

An example for IPv6 would be if a fragment is being added and is at
the stage of being inserted in the hash after the hash lock is
released, but before inet_frag_lru_add executes (or is able to obtain
the lru lock) another overlapping fragment for the same flow arrives
at a different CPU which finds it in the hash, but since it's
overlapping it drops it invoking inet_frag_kill and thus removing all
guarding refcounts, and afterwards freeing it by invoking
inet_frag_put which removes the last refcount added previously by
inet_frag_find, then inet_frag_lru_add gets executed by
inet_frag_intern and we have a freed fragment in the lru_list.

The fix is simple, just move the lru_add under the hash chain locked
region so when a removing function is called it'll have to wait for
the fragment to be added to the lru_list, and then it'll remove it (it
works because the hash chain removal is done before the lru_list one
and there's no window between the two list adds when the frag can get
dropped). With this fix applied I couldn't kill the same machine in 24
hours with the same setup.

Fixes: 3ef0eb0d ("net: frag, move LRU list maintenance outside of
rwlock")

CC: Florian Westphal <fw@strlen.de>
CC: Jesper Dangaard Brouer <brouer@redhat.com>
CC: David S. Miller <davem@davemloft.net>
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Acked-by: NJesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull networking fixes from David Miller:

 1) Fix memory leak in ieee80211_prep_connection(), sta_info leaked on
    error.  From Eytan Lifshitz.

 2) Unintentional switch case fallthrough in nft_reject_inet_eval(),
    from Patrick McHardy.

 3) Must check if payload lenth is a power of 2 in
    nft_payload_select_ops(), from Nikolay Aleksandrov.

 4) Fix mis-checksumming in xen-netfront driver, ip_hdr() is not in the
    correct place when we invoke skb_checksum_setup().  From Wei Liu.

 5) TUN driver should not advertise HW vlan offload features in
    vlan_features.  Fix from Fernando Luis Vazquez Cao.

 6) IPV6_VTI needs to select NET_IPV_TUNNEL to avoid build errors, fix
    from Steffen Klassert.

 7) Add missing locking in xfrm_migrade_state_find(), we must hold the
    per-namespace xfrm_state_lock while traversing the lists.  Fix from
    Steffen Klassert.

 8) Missing locking in ath9k driver, access to tid->sched must be done
    under ath_txq_lock().  Fix from Stanislaw Gruszka.

 9) Fix two bugs in TCP fastopen.  First respect the size argument given
    to tcp_sendmsg() in the fastopen path, and secondly prevent
    tcp_send_syn_data() from potentially using order-5 allocations.
    From Eric Dumazet.

10) Fix handling of default neigh garbage collection params, from Jiri
    Pirko.

11) Fix cwnd bloat and over-inflation of RTT when transmit segmentation
    is in use.  From Eric Dumazet.

12) Missing initialization of Realtek r8169 driver's statistics
    seqlocks.  Fix from Kyle McMartin.

13) Fix RTNL assertion failures in 802.3ad and AB ARP monitor of bonding
    driver, from Ding Tianhong.

14) Bonding slave release race can cause divide by zero, fix from
    Nikolay Aleksandrov.

15) Overzealous return from neigh_periodic_work() causes reachability
    time to not be computed.  Fix from Duain Jiong.

16) Fix regression in ipv6_find_hdr(), it should not return -ENOENT when
    a specific target is specified and found.  From Hans Schillstrom.

17) Fix VLAN tag stripping regression in BNA driver, from Ivan Vecera.

18) Tail loss probe can calculate bogus RTTs due to missing packet
    marking on retransmit.  Fix from Yuchung Cheng.

19) We cannot do skb_dst_drop() in iptunnel_pull_header() because
    multicast loopback detection in later code paths need access to
    skb_rtable().  Fix from Xin Long.

20) The macvlan driver regresses in that it propagates lower device
    offload support disables into itself, causing severe slowdowns when
    running over a bridge.  Provide the software offloads always on
    macvlan devices to deal with this and the regression is gone.  From
    Vlad Yasevich.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (103 commits)
  macvlan: Add support for 'always_on' offload features
  net: sctp: fix sctp_sf_do_5_1D_ce to verify if we/peer is AUTH capable
  ip_tunnel:multicast process cause panic due to skb->_skb_refdst NULL pointer
  net: cpsw: fix cpdma rx descriptor leak on down interface
  be2net: isolate TX workarounds not applicable to Skyhawk-R
  be2net: Fix skb double free in be_xmit_wrokarounds() failure path
  be2net: clear promiscuous bits in adapter->flags while disabling promiscuous mode
  be2net: Fix to reset transparent vlan tagging
  qlcnic: dcb: a couple off by one bugs
  tcp: fix bogus RTT on special retransmission
  hsr: off by one sanity check in hsr_register_frame_in()
  can: remove CAN FD compatibility for CAN 2.0 sockets
  can: flexcan: factor out soft reset into seperate funtion
  can: flexcan: flexcan_remove(): add missing netif_napi_del()
  can: flexcan: fix transition from and to freeze mode in chip_{,un}freeze
  can: flexcan: factor out transceiver {en,dis}able into seperate functions
  can: flexcan: fix transition from and to low power mode in chip_{en,dis}able
  can: flexcan: flexcan_open(): fix error path if flexcan_chip_start() fails
  can: flexcan: fix shutdown: first disable chip, then all interrupts
  USB AX88179/178A: Support D-Link DUB-1312
  ...

Pull regulator fixes from Mark Brown:
 "A couple of fixes here which ensure that regulators using the core
  support for GPIO enables work in all cases by ensuring that helpers
  are used consistently rather than open coding in places and hence not
  having GPIO support in some of them"

* tag 'regulator-v3.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: core: Replace direct ops->disable usage
  regulator: core: Replace direct ops->enable usage

Merge misc fixes from Andrew Morton.

* emailed patches from Andrew Morton akpm@linux-foundation.org>:
  mm: page_alloc: exempt GFP_THISNODE allocations from zone fairness
  mm: numa: bugfix for LAST_CPUPID_NOT_IN_PAGE_FLAGS
  MAINTAINERS: add and correct types of some "T:" entries
  MAINTAINERS: use tab for separator
  rapidio/tsi721: fix tasklet termination in dma channel release
  hfsplus: fix remount issue
  zram: avoid null access when fail to alloc meta
  sh: prefix sh-specific "CCR" and "CCR2" by "SH_"
  ocfs2: fix quota file corruption
  drivers/rtc/rtc-s3c.c: fix incorrect way of save/restore of S3C2410_TICNT for TYPE_S3C64XX
  kallsyms: fix absolute addresses for kASLR
  scripts/gen_initramfs_list.sh: fix flags for initramfs LZ4 compression
  mm: include VM_MIXEDMAP flag in the VM_SPECIAL list to avoid m(un)locking
  memcg: reparent charges of children before processing parent
  memcg: fix endless loop in __mem_cgroup_iter_next()
  lib/radix-tree.c: swapoff tmpfs radix_tree: remember to rcu_read_unlock
  dma debug: account for cachelines and read-only mappings in overlap tracking
  mm: close PageTail race
  MAINTAINERS: EDAC: add Mauro and Borislav as interim patch collectors

Jan Stancek reports manual page migration encountering allocation
failures after some pages when there is still plenty of memory free, and
bisected the problem down to commit 81c0a2bb ("mm: page_alloc: fair
zone allocator policy").

The problem is that GFP_THISNODE obeys the zone fairness allocation
batches on one hand, but doesn't reset them and wake kswapd on the other
hand.  After a few of those allocations, the batches are exhausted and
the allocations fail.

Fixing this means either having GFP_THISNODE wake up kswapd, or
GFP_THISNODE not participating in zone fairness at all.  The latter
seems safer as an acute bugfix, we can clean up later.
Reported-by: NJan Stancek <jstancek@redhat.com>
Signed-off-by: NJohannes Weiner <hannes@cmpxchg.org>
Acked-by: NRik van Riel <riel@redhat.com>
Acked-by: NMel Gorman <mgorman@suse.de>
Cc: <stable@kernel.org>		[3.12+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When doing some numa tests on powerpc, I triggered an oops bug.  I find
it is caused by using page->_last_cpupid.  It should be initialized as
"-1 & LAST_CPUPID_MASK", but not "-1".  Otherwise, in task_numa_fault(),
we will miss the checking (last_cpupid == (-1 & LAST_CPUPID_MASK)).  And
finally cause an oops bug in task_numa_group(), since the online cpu is
less than possible cpu.  This happen with CONFIG_SPARSE_VMEMMAP disabled

Call trace:

  SMP NR_CPUS=64 NUMA PowerNV
  Modules linked in:
  CPU: 24 PID: 804 Comm: systemd-udevd Not tainted3.13.0-rc1+ #32
  task: c000001e2746aa80 ti: c000001e32c50000 task.ti:c000001e32c50000
  REGS: c000001e32c53510 TRAP: 0300   Not tainted(3.13.0-rc1+)
  MSR: 9000000000009032 <SF,HV,EE,ME,IR,DR,RI>  CR:28024424  XER: 20000000
  CFAR: c000000000009324 DAR: 7265717569726857 DSISR:40000000 SOFTE: 1
  NIP  .task_numa_fault+0x1470/0x2370
  LR  .task_numa_fault+0x1468/0x2370
  Call Trace:
   .task_numa_fault+0x1468/0x2370 (unreliable)
   .do_numa_page+0x480/0x4a0
   .handle_mm_fault+0x4ec/0xc90
   .do_page_fault+0x3a8/0x890
   handle_page_fault+0x10/0x30
  Instruction dump:
  3c82fefb 3884b138 48d9cff1 60000000 48000574 3c62fefb3863af78 3c82fefb
  3884b138 48d9cfd5 60000000 e93f0100 <812902e4> 7d2907b45529063e 7d2a07b4
  ---[ end trace 15f2510da5ae07cf ]---
Signed-off-by: NLiu Ping Fan <pingfank@linux.vnet.ibm.com>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: NPeter Zijlstra <peterz@infradead.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Tree location entries should start with the appropriate type.

Add git to some, hg to another.

Neaten tree type description.
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Convert whitespace to single tab for separators.
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch is a modification of the patch originally proposed by
Xiaotian Feng <xtfeng@gmail.com>: https://lkml.org/lkml/2012/11/5/413
This new version disables DMA channel interrupts and ensures that the
tasklet wil not be scheduled again before calling tasklet_kill().

Unfortunately the updated patch was not released at that time due to
planned rework of Tsi721 mport driver to use threaded interrupts (which
has yet to happen).  Recently the issue was reported again:
https://lkml.org/lkml/2014/2/19/762.

Description from the original Xiaotian's patch:

 "Some drivers use tasklet_disable in device remove/release process,
  tasklet_disable will inc tasklet->count and return.  If the tasklet is
  not handled yet under some softirq pressure, the tasklet will be
  placed on the tasklet_vec, never have a chance to be excuted.  This
  might lead to a heavy loaded ksoftirqd, wakeup with pending_softirq,
  but tasklet is disabled.  tasklet_kill should be used in this case."

This patch is applicable to kernel versions starting from v3.5.
Signed-off-by: NAlexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Xiaotian Feng <xtfeng@gmail.com>
Reviewed-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Mike Galbraith <bitbucket@online.de>
Cc: <stable@vger.kernel.org>	[3.5+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Current implementation of HFS+ driver has small issue with remount
option.  Namely, for example, you are unable to remount from RO mode
into RW mode by means of command "mount -o remount,rw /dev/loop0
/mnt/hfsplus".  Trying to execute sequence of commands results in an
error message:

  mount /dev/loop0 /mnt/hfsplus
  mount -o remount,ro /dev/loop0 /mnt/hfsplus
  mount -o remount,rw /dev/loop0 /mnt/hfsplus

  mount: you must specify the filesystem type

  mount -t hfsplus -o remount,rw /dev/loop0 /mnt/hfsplus

  mount: /mnt/hfsplus not mounted or bad option

The reason of such issue is failure of mount syscall:

  mount("/dev/loop0", "/mnt/hfsplus", 0x2282a60, MS_MGC_VAL|MS_REMOUNT, NULL) = -1 EINVAL (Invalid argument)

Namely, hfsplus_parse_options_remount() method receives empty "input"
argument and return false in such case.  As a result, hfsplus_remount()
returns -EINVAL error code.

This patch fixes the issue by means of return true for the case of empty
"input" argument in hfsplus_parse_options_remount() method.
Signed-off-by: NVyacheslav Dubeyko <slava@dubeyko.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

zram_meta_alloc could fail so caller should check it.  Otherwise, your
system will hang.
Signed-off-by: NMinchan Kim <minchan@kernel.org>
Acked-by: NJerome Marchand <jmarchan@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit bcf24e1d ("mmc: omap_hsmmc: use the generic config for
omap2plus devices"), enabled the build for other platforms for compile
testing.

sh-allmodconfig now fails with:

    include/linux/omap-dma.h:171:8: error: expected identifier before numeric constant
    make[4]: *** [drivers/mmc/host/omap_hsmmc.o] Error 1

This happens because SuperH #defines "CCR", which is one of the enum
values in include/linux/omap-dma.h.  There's a similar issue with "CCR2"
on sh2a.

As "CCR" and "CCR2" are too generic names for global #defines, prefix
them with "SH_" to fix this.
Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Global quota files are accessed from different nodes.  Thus we cannot
cache offset of quota structure in the quota file after we drop our node
reference count to it because after that moment quota structure may be
freed and reallocated elsewhere by a different node resulting in
corruption of quota file.

Fix the problem by clearing dq_off when we are releasing dquot structure.
We also remove the DB_READ_B handling because it is useless -
DQ_ACTIVE_B is set iff DQ_READ_B is set.
Signed-off-by: NJan Kara <jack@suse.cz>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Reviewed-by: NMark Fasheh <mfasheh@suse.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

On exynos5250, exynos5420 and exynos5260 it was observed that, after 1
cycle of S2R, the rtc-tick occurs at a very fast rate as compared to the
rtc-tick occuring before S2R.

This patch fixes the above issue by correcting the wrong way of
save/restore of S3C2410_TICNT for TYPE_S3C64XX.
Signed-off-by: NVikas Sajjan <vikas.sajjan@samsung.com>
Cc: Grant Likely <grant.likely@linaro.org>
Cc: Rob Herring <robh+dt@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Currently symbols that are absolute addresses are incorrectly displayed
in /proc/kallsyms if the kernel is loaded with kASLR.

The problem was that the scripts/kallsyms.c file which generates the
array of symbol names and addresses uses an relocatable value for all
symbols, even absolute symbols.  This patch fixes that.

Several kallsyms output in different boot states for comparison:

  $ egrep '_(stext|_per_cpu_(start|end))' /root/kallsyms.nokaslr
  0000000000000000 D __per_cpu_start
  0000000000014280 D __per_cpu_end
  ffffffff810001c8 T _stext
  $ egrep '_(stext|_per_cpu_(start|end))' /root/kallsyms.kaslr1
  000000001f200000 D __per_cpu_start
  000000001f214280 D __per_cpu_end
  ffffffffa02001c8 T _stext
  $ egrep '_(stext|_per_cpu_(start|end))' /root/kallsyms.kaslr2
  000000000d400000 D __per_cpu_start
  000000000d414280 D __per_cpu_end
  ffffffff8e4001c8 T _stext
  $ egrep '_(stext|_per_cpu_(start|end))' /root/kallsyms.kaslr-fixed
  0000000000000000 D __per_cpu_start
  0000000000014280 D __per_cpu_end
  ffffffffadc001c8 T _stext
Signed-off-by: NAndy Honig <ahonig@google.com>
Signed-off-by: NKees Cook <keescook@chromium.org>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

LZ4 as implemented in the kernel differs from the default method now
used by the reference implementation of LZ4.  Until the in-kernel method
is updated to support the new default, passing the legacy flag (-l) to
the compressor is necessary.  Without this flag the kernel-generated,
LZ4-compressed initramfs is junk.

Kyungsik said:

: It seems that lz4 supports legacy format with the same option as lz4c
: does.  Just looking at the first few bytes of lz4 compressed image, we can
: see whether it is new format or not.
:
: It shows new format magic number without this patch.  New format magic
: number is 0x184d2204.
:
: $ hexdump -C ./initramfs_data.cpio.lz4 |more
: 00000000  04 22 4d 18 64 70 b9 69 (Little Endian)
: ...
:
: Currently kernel supports legacy format only.
Signed-off-by: NDaniel M. Weeks <dan@danweeks.net>
Cc: Michal Marek <mmarek@suse.cz>
Acked-by: NKyungsik Lee <kyungsik.lee@lge.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Daniel Borkmann reported a VM_BUG_ON assertion failing:

  ------------[ cut here ]------------
  kernel BUG at mm/mlock.c:528!
  invalid opcode: 0000 [#1] SMP
  Modules linked in: ccm arc4 iwldvm [...]
   video
  CPU: 3 PID: 2266 Comm: netsniff-ng Not tainted 3.14.0-rc2+ #8
  Hardware name: LENOVO 2429BP3/2429BP3, BIOS G4ET37WW (1.12 ) 05/29/2012
  task: ffff8801f87f9820 ti: ffff88002cb44000 task.ti: ffff88002cb44000
  RIP: 0010:[<ffffffff81171ad0>]  [<ffffffff81171ad0>] munlock_vma_pages_range+0x2e0/0x2f0
  Call Trace:
    do_munmap+0x18f/0x3b0
    vm_munmap+0x41/0x60
    SyS_munmap+0x22/0x30
    system_call_fastpath+0x1a/0x1f
  RIP   munlock_vma_pages_range+0x2e0/0x2f0
  ---[ end trace a0088dcf07ae10f2 ]---

because munlock_vma_pages_range() thinks it's unexpectedly in the middle
of a THP page.  This can be reproduced with default config since 3.11
kernels.  A reproducer can be found in the kernel's selftest directory
for networking by running ./psock_tpacket.

The problem is that an order=2 compound page (allocated by
alloc_one_pg_vec_page() is part of the munlocked VM_MIXEDMAP vma (mapped
by packet_mmap()) and mistaken for a THP page and assumed to be order=9.

The checks for THP in munlock came with commit ff6a6da6 ("mm:
accelerate munlock() treatment of THP pages"), i.e.  since 3.9, but did
not trigger a bug.  It just makes munlock_vma_pages_range() skip such
compound pages until the next 512-pages-aligned page, when it encounters
a head page.  This is however not a problem for vma's where mlocking has
no effect anyway, but it can distort the accounting.

Since commit 7225522b ("mm: munlock: batch non-THP page isolation
and munlock+putback using pagevec") this can trigger a VM_BUG_ON in
PageTransHuge() check.

This patch fixes the issue by adding VM_MIXEDMAP flag to VM_SPECIAL, a
list of flags that make vma's non-mlockable and non-mergeable.  The
reasoning is that VM_MIXEDMAP vma's are similar to VM_PFNMAP, which is
already on the VM_SPECIAL list, and both are intended for non-LRU pages
where mlocking makes no sense anyway.  Related Lkml discussion can be
found in [2].

 [1] tools/testing/selftests/net/psock_tpacket
 [2] https://lkml.org/lkml/2014/1/10/427Signed-off-by: NVlastimil Babka <vbabka@suse.cz>
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Reported-by: NDaniel Borkmann <dborkman@redhat.com>
Tested-by: NDaniel Borkmann <dborkman@redhat.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: John David Anglin <dave.anglin@bell.net>
Cc: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Tested-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: NRik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org> [3.11.x+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Sometimes the cleanup after memcg hierarchy testing gets stuck in
mem_cgroup_reparent_charges(), unable to bring non-kmem usage down to 0.

There may turn out to be several causes, but a major cause is this: the
workitem to offline parent can get run before workitem to offline child;
parent's mem_cgroup_reparent_charges() circles around waiting for the
child's pages to be reparented to its lrus, but it's holding
cgroup_mutex which prevents the child from reaching its
mem_cgroup_reparent_charges().

Further testing showed that an ordered workqueue for cgroup_destroy_wq
is not always good enough: percpu_ref_kill_and_confirm's call_rcu_sched
stage on the way can mess up the order before reaching the workqueue.

Instead, when offlining a memcg, call mem_cgroup_reparent_charges() on
all its children (and grandchildren, in the correct order) to have their
charges reparented first.

Fixes: e5fca243 ("cgroup: use a dedicated workqueue for cgroup destruction")
Signed-off-by: NFilipe Brandenburger <filbranden@google.com>
Signed-off-by: NHugh Dickins <hughd@google.com>
Reviewed-by: NTejun Heo <tj@kernel.org>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: <stable@vger.kernel.org>	[v3.10+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit 0eef6156 ("memcg: fix css reference leak and endless loop in
mem_cgroup_iter") got the interaction with the commit a few before it
d8ad3055 ("mm/memcg: iteration skip memcgs not yet fully
initialized") slightly wrong, and we didn't notice at the time.

It's elusive, and harder to get than the original, but for a couple of
days before rc1, I several times saw a endless loop similar to that
supposedly being fixed.

This time it was a tighter loop in __mem_cgroup_iter_next(): because we
can get here when our root has already been offlined, and the ordering
of conditions was such that we then just cycled around forever.

Fixes: 0eef6156 ("memcg: fix css reference leak and endless loop in mem_cgroup_iter").
Signed-off-by: NHugh Dickins <hughd@google.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: <stable@vger.kernel.org>	[3.12+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Running fsx on tmpfs with concurrent memhog-swapoff-swapon, lots of

  BUG: sleeping function called from invalid context at kernel/fork.c:606
  in_atomic(): 0, irqs_disabled(): 0, pid: 1394, name: swapoff
  1 lock held by swapoff/1394:
   #0:  (rcu_read_lock){.+.+.+}, at: [<ffffffff812520a1>] radix_tree_locate_item+0x1f/0x2b6

followed by

  ================================================
  [ BUG: lock held when returning to user space! ]
  3.14.0-rc1 #3 Not tainted
  ------------------------------------------------
  swapoff/1394 is leaving the kernel with locks still held!
  1 lock held by swapoff/1394:
   #0:  (rcu_read_lock){.+.+.+}, at: [<ffffffff812520a1>] radix_tree_locate_item+0x1f/0x2b6

after which the system recovered nicely.

Whoops, I long ago forgot the rcu_read_unlock() on one unlikely branch.

Fixes e504f3fd ("tmpfs radix_tree: locate_item to speed up swapoff")
Signed-off-by: NHugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>