Skip to content

L
linux
Switch branch/tag
  1. 07 Feb, 2014 1 commit
    • Will Deacon's avatar
      arm64: atomics: fix use of acquire + release for full barrier semantics · 8e86f0b4
      Will Deacon authored 
      Linux requires a number of atomic operations to provide full barrier
semantics, that is no memory accesses after the operation can be
observed before any accesses up to and including the operation in
program order.

On arm64, these operations have been incorrectly implemented as follows:

	// A, B, C are independent memory locations

	<Access [A]>

	// atomic_op (B)
1:	ldaxr	x0, [B]		// Exclusive load with acquire
	<op(B)>
	stlxr	w1, x0, [B]	// Exclusive store with release
	cbnz	w1, 1b

	<Access [C]>

The assumption here being that two half barriers are equivalent to a
full barrier, so the only permitted ordering would be A -> B -> C
(where B is the atomic operation involving both a load and a store).

Unfortunately, this is not the case by the letter of the architecture
and, in fact, the accesses to A and C are permitted to pass their
nearest half barrier resulting in orderings such as Bl -> A -> C -> Bs
or Bl -> C -> A -> Bs (where Bl is the load-acquire on B and Bs is the
store-release on B). This is a clear violation of the full barrier
requirement.

The simple way to fix this is to implement the same algorithm as ARMv7
using explicit barriers:

	<Access [A]>

	// atomic_op (B)
	dmb	ish		// Full barrier
1:	ldxr	x0, [B]		// Exclusive load
	<op(B)>
	stxr	w1, x0, [B]	// Exclusive store
	cbnz	w1, 1b
	dmb	ish		// Full barrier

	<Access [C]>

but this has the undesirable effect of introducing *two* full barrier
instructions. A better approach is actually the following, non-intuitive
sequence:

	<Access [A]>

	// atomic_op (B)
1:	ldxr	x0, [B]		// Exclusive load
	<op(B)>
	stlxr	w1, x0, [B]	// Exclusive store with release
	cbnz	w1, 1b
	dmb	ish		// Full barrier

	<Access [C]>

The simple observations here are:

  - The dmb ensures that no subsequent accesses (e.g. the access to C)
    can enter or pass the atomic sequence.

  - The dmb also ensures that no prior accesses (e.g. the access to A)
    can pass the atomic sequence.

  - Therefore, no prior access can pass a subsequent access, or
    vice-versa (i.e. A is strictly ordered before C).

  - The stlxr ensures that no prior access can pass the store component
    of the atomic operation.

The only tricky part remaining is the ordering between the ldxr and the
access to A, since the absence of the first dmb means that we're now
permitting re-ordering between the ldxr and any prior accesses.

From an (arbitrary) observer's point of view, there are two scenarios:

  1. We have observed the ldxr. This means that if we perform a store to
     [B], the ldxr will still return older data. If we can observe the
     ldxr, then we can potentially observe the permitted re-ordering
     with the access to A, which is clearly an issue when compared to
     the dmb variant of the code. Thankfully, the exclusive monitor will
     save us here since it will be cleared as a result of the store and
     the ldxr will retry. Notice that any use of a later memory
     observation to imply observation of the ldxr will also imply
     observation of the access to A, since the stlxr/dmb ensure strict
     ordering.

  2. We have not observed the ldxr. This means we can perform a store
     and influence the later ldxr. However, that doesn't actually tell
     us anything about the access to [A], so we've not lost anything
     here either when compared to the dmb variant.

This patch implements this solution for our barriered atomic operations,
ensuring that we satisfy the full barrier requirements where they are
needed.

Cc: <stable@vger.kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: default avatarWill Deacon <will.deacon@arm.com>
Signed-off-by: default avatarCatalin Marinas <catalin.marinas@arm.com>
      8e86f0b4
  3. 06 Feb, 2014 1 commit
  5. 05 Feb, 2014 9 commits
  7. 04 Feb, 2014 1 commit
  9. 03 Feb, 2014 4 commits
    • Linus Torvalds's avatar
      Linus 3.14-rc1 · 38dbfb59
      Linus Torvalds authored
      38dbfb59
    • Linus Torvalds's avatar
      Merge branch 'parisc-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux · 69048e01
      Linus Torvalds authored 
      Pull parisc updates from Helge Deller:
 "The three major changes in this patchset is a implementation for
  flexible userspace memory maps, cache-flushing fixes (again), and a
  long-discussed ABI change to make EWOULDBLOCK the same value as
  EAGAIN.

  parisc has been the only platform where we had EWOULDBLOCK != EAGAIN
  to keep HP-UX compatibility.  Since we will probably never implement
  full HP-UX support, we prefer to drop this compatibility to make it
  easier for us with Linux userspace programs which mostly never checked
  for both values.  We don't expect major fall-outs because of this
  change, and if we face some, we will simply rebuild the necessary
  applications in the debian archives"

* 'parisc-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: add flexible mmap memory layout support
  parisc: Make EWOULDBLOCK be equal to EAGAIN on parisc
  parisc: convert uapi/asm/stat.h to use native types only
  parisc: wire up sched_setattr and sched_getattr
  parisc: fix cache-flushing
  parisc/sti_console: prefer Linux fonts over built-in ROM fonts
      69048e01
    • Mikulas Patocka's avatar
      hpfs: optimize quad buffer loading · 1c0b8a7a
      Mikulas Patocka authored 
      HPFS needs to load 4 consecutive 512-byte sectors when accessing the
directory nodes or bitmaps.  We can't switch to 2048-byte block size
because files are allocated in the units of 512-byte sectors.

Previously, the driver would allocate a 2048-byte area using kmalloc,
copy the data from four buffers to this area and eventually copy them
back if they were modified.

In the current implementation of the buffer cache, buffers are allocated
in the pagecache.  That means that 4 consecutive 512-byte buffers are
stored in consecutive areas in the kernel address space.  So, we don't
need to allocate extra memory and copy the content of the buffers there.

This patch optimizes the code to avoid copying the buffers.  It checks
if the four buffers are stored in contiguous memory - if they are not,
it falls back to allocating a 2048-byte area and copying data there.
Signed-off-by: default avatarMikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
      1c0b8a7a
    • Mikulas Patocka's avatar
      hpfs: remember free space · 2cbe5c76
      Mikulas Patocka authored 
      Previously, hpfs scanned all bitmaps each time the user asked for free
space using statfs.  This patch changes it so that hpfs scans the
bitmaps only once, remembes the free space and on next invocation of
statfs it returns the value instantly.

New versions of wine are hammering on the statfs syscall very heavily,
making some games unplayable when they're stored on hpfs, with load
times in minutes.

This should be backported to the stable kernels because it fixes
user-visible problem (excessive level load times in wine).
Signed-off-by: default avatarMikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Cc: stable@vger.kernel.org
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
      2cbe5c76
  11. 02 Feb, 2014 12 commits
  13. 01 Feb, 2014 12 commits
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7