do_blockdev_direct_IO() increments and decrements the inode
->i_dio_count for each IO operation. It does this to protect against
truncate of a file. Block devices don't need this sort of protection.
For a capable multiqueue setup, this atomic int is the only shared
state between applications accessing the device for O_DIRECT, and it
presents a scaling wall for that. In my testing, as much as 30% of
system time is spent incrementing and decrementing this value. A mixed
read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with
better latencies too. Before:
clat percentiles (usec):
| 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34],
| 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35],
| 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80],
| 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155],
| 99.99th=[ 165]
After:
clat percentiles (usec):
| 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149],
| 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171],
| 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270],
| 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422],
| 99.99th=[ 438]
In other setups, Robert Elliott reported seeing good performance
improvements:
https://lkml.org/lkml/2015/4/3/557
The more applications accessing the device, the worse it gets.
Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells
do_blockdev_direct_IO() that it need not worry about incrementing
or decrementing the inode i_dio_count for this caller.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Elliott, Robert (Server Storage) <elliott@hp.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jens Axboe <axboe@fb.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
that's the bulk of filesystem drivers dealing with inodes of their own
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Impose ordering on accesses of d_inode and d_flags to avoid the need to do
this:
if (!dentry->d_inode || d_is_negative(dentry)) {
when this:
if (d_is_negative(dentry)) {
should suffice.
This check is especially problematic if a dentry can have its type field set
to something other than DENTRY_MISS_TYPE when d_inode is NULL (as in
unionmount).
What we really need to do is stick a write barrier between setting d_inode and
setting d_flags and a read barrier between reading d_flags and reading
d_inode.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Supply two functions to test whether a filesystem's own dentries are positive
or negative (d_really_is_positive() and d_really_is_negative()).
The problem is that the DCACHE_ENTRY_TYPE field of dentry->d_flags may be
overridden by the union part of a layered filesystem and isn't thus
necessarily indicative of the type of dentry.
Normally, this would involve a negative dentry (ie. ->d_inode == NULL) having
->d_layer.lower pointed to a lower layer dentry, DCACHE_PINNING_LOWER set and
the DCACHE_ENTRY_TYPE field set to something other than DCACHE_MISS_TYPE - but
it could also involve, say, a DCACHE_SPECIAL_TYPE being overridden to
DCACHE_WHITEOUT_TYPE if a 0,0 chardev is detected in the top layer.
However, inside a filesystem, when that fs is looking at its own dentries, it
probably wants to know if they are really negative or not - and doesn't care
about the fallthrough bits used by the union.
To this end, a filesystem should normally use d_really_is_positive/negative()
when looking at its own dentries rather than d_is_positive/negative() and
should use d_inode() to get at the inode.
Anyone looking at someone else's dentries (this includes pathwalk) should use
d_is_xxx() and d_backing_inode().
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... returning -E... upon error and amount of data left in iter after
(possible) truncation upon success. Note, that normal case gives
a non-zero (positive) return value, so any tests for != 0 _must_ be
updated.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Conflicts:
fs/ext4/file.c
Most filesystems call through to these at some point, so we'll start
here.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
all remaining instances of aio_{read,write} (all 4 of them) have explicit
->read and ->write resp.; do_sync_read/do_sync_write is never called by
__vfs_read/__vfs_write anymore and no other users had been left.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
All places outside of core VFS that checked ->read and ->write for being NULL or
called the methods directly are gone now, so NULL {read,write} with non-NULL
{read,write}_iter will do the right thing in all cases.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
simillar to iov_iter_fault_in_readable() but differs in that it is
not limited to faulting in the first iovec and instead faults in
"bytes" bytes iterating over the iovecs as necessary.
Also, instead of only faulting in the first and last page of the
range, all pages are faulted in.
This function is needed by NTFS when it does multi page file
writes.
Signed-off-by: Anton Altaparmakov <anton@tuxera.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
just make const char iname[] the last member and compare name->name with
name->iname instead of checking name->separate
We need to make sure that out-of-line name doesn't end up allocated adjacent
to struct filename refering to it; fortunately, it's easy to achieve - just
allocate that struct filename with one byte in ->iname[], so that ->iname[0]
will be inside the same object and thus have an address different from that
of out-of-line name [spotted by Boqun Feng <boqun.feng@gmail.com>]
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
trivial conflict in net/socket.c and non-trivial one in crypto -
that one had evaded aio_complete() removal.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Johan Hedberg says:
====================
pull request: bluetooth-next 2015-04-04
Here's what's probably the last bluetooth-next pull request for 4.1:
- Fixes for LE advertising data & advertising parameters
- Fix for race condition with HCI_RESET flag
- New BNEPGETSUPPFEAT ioctl, needed for certification
- New HCI request callback type to get the resulting skb
- Cleanups to use BIT() macro wherever possible
- Consolidate Broadcom device entries in the btusb HCI driver
- Check for valid flags in CMTP, HIDP & BNEP
- Disallow local privacy & OOB data combo to prevent a potential race
- Expose SMP & ECDH selftest results through debugfs
- Expose current Device ID info through debugfs
Please let me know if there are any issues pulling. Thanks.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/mellanox/mlx4/cmd.c
net/core/fib_rules.c
net/ipv4/fib_frontend.c
The fib_rules.c and fib_frontend.c conflicts were locking adjustments
in 'net' overlapping addition and removal of code in 'net-next'.
The mlx4 conflict was a bug fix in 'net' happening in the same
place a constant was being replaced with a more suitable macro.
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
1) In TCP, don't register an FRTO for cumulatively ACK'd data that was
previously SACK'd, from Neal Cardwell.
2) Need to hold RNL mutex in ipv4 multicast code namespace cleanup,
from Cong WANG.
3) Similarly we have to hold RNL mutex for fib_rules_unregister(), also
from Cong WANG.
4) Revert and rework netns nsid allocation fix, from Nicolas Dichtel.
5) When we encapsulate for a tunnel device, skb->sk still points to the
user socket. So this leads to cases where we retraverse the
ipv4/ipv6 output path with skb->sk being of some other address
family (f.e. AF_PACKET). This can cause things to crash since the
ipv4 output path is dereferencing an AF_PACKET socket as if it were
an ipv4 one.
The short term fix for 'net' and -stable is to elide these socket
checks once we've entered an encapsulation sequence by testing
xmit_recursion.
Longer term we have a better solution wherein we pass the tunnel's
socket down through the output paths, but that is way too invasive
for 'net' and -stable.
From Hannes Frederic Sowa.
6) l2tp_init() failure path forgets to unregister per-net ops, from
Cong WANG.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
net/mlx4_core: Fix error message deprecation for ConnectX-2 cards
net: dsa: fix filling routing table from OF description
l2tp: unregister l2tp_net_ops on failure path
mvneta: dont call mvneta_adjust_link() manually
ipv6: protect skb->sk accesses from recursive dereference inside the stack
netns: don't allocate an id for dead netns
Revert "netns: don't clear nsid too early on removal"
ip6mr: call del_timer_sync() in ip6mr_free_table()
net: move fib_rules_unregister() under rtnl lock
ipv4: take rtnl_lock and mark mrt table as freed on namespace cleanup
tcp: fix FRTO undo on cumulative ACK of SACKed range
xen-netfront: transmit fully GSO-sized packets
teach ->mremap() method to return an error and have it fail for
aio mappings in process of being killed
Note that in case of ->mremap() failure we need to undo move_page_tables()
we'd already done; we could call ->mremap() first, but then the failure of
move_page_tables() would require undoing whatever _successful_ ->mremap()
has done, which would be a lot more headache in general.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Commit 608cd71a9c ("tc: bpf: generalize pedit action") has added the
possibility to mangle packet data to BPF programs in the tc pipeline.
This patch adds two helpers bpf_l3_csum_replace() and bpf_l4_csum_replace()
for fixing up the protocol checksums after the packet mangling.
It also adds 'flags' argument to bpf_skb_store_bytes() helper to avoid
unnecessary checksum recomputations when BPF programs adjusting l3/l4
checksums and documents all three helpers in uapi header.
Moreover, a sample program is added to show how BPF programs can make use
of the mangle and csum helpers.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
We should not consult skb->sk for output decisions in xmit recursion
levels > 0 in the stack. Otherwise local socket settings could influence
the result of e.g. tunnel encapsulation process.
ipv6 does not conform with this in three places:
1) ip6_fragment: we do consult ipv6_npinfo for frag_size
2) sk_mc_loop in ipv6 uses skb->sk and checks if we should
loop the packet back to the local socket
3) ip6_skb_dst_mtu could query the settings from the user socket and
force a wrong MTU
Furthermore:
In sk_mc_loop we could potentially land in WARN_ON(1) if we use a
PF_PACKET socket ontop of an IPv6-backed vxlan device.
Reuse xmit_recursion as we are currently only interested in protecting
tunnel devices.
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
That way we don't have to reinstantiate another nf_hook_state
on the stack of the nf_reinject() path.
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of passing a large number of arguments down into the nf_hook()
entry points, create a structure which carries this state down through
the hook processing layers.
This makes is so that if we want to change the types or signatures of
any of these pieces of state, there are less places that need to be
changed.
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull input subsystem fixes from Dmitry Torokhov:
"A fix for ALPS driver for issue introduced in the latest update and a
tweak for yet another Lenovo box in Synaptics.
There will be more ALPS tweaks coming.."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Input: define INPUT_PROP_ACCELEROMETER behavior
Input: synaptics - fix min-max quirk value for E440
Input: synaptics - add quirk for Thinkpad E440
Input: ALPS - fix max coordinates for v5 and v7 protocols
Input: add MT_TOOL_PALM
Pull block layer fix from Jens Axboe:
"Just one patch in this pull request, fixing a regression caused by a
'mathematically correct' change to lcm()"
* 'for-linus' of git://git.kernel.dk/linux-block:
block: fix blk_stack_limits() regression due to lcm() change
Currently fixed_phy uses a callback to periodically poll the link state.
This patch adds the fixed_phy_update_state() API.
It solves the following problems:
- On link state interrupt, MAC driver can't update status.
Instead it needs to provide the callback to periodically query
the HW about the link state. It is more efficient to update status
after interrupt.
- The callback needs to be unregistered before phy_disconnect(),
or otherwise it will be called with net_dev==NULL. phy_disconnect()
does not have enough info to unregister the callback automatically.
- The callback needs to be registered before of_phy_connect() to
avoid running with outdated state, but of_phy_connect() returns the
phy_device pointer, which is needed to register the callback. Registering
it before of_phy_connect() will therefore require a hack to get the
pointer earlier.
Overall, this addition makes the subsequent patch that implements
SGMII link status for mvneta, much cleaner.
CC: Florian Fainelli <f.fainelli@gmail.com>
CC: netdev@vger.kernel.org
CC: linux-kernel@vger.kernel.org
Signed-off-by: Stas Sergeev <stsp@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
