[NaCl SDK] Upate atomic ops in nacl_io

[NaCl SDK] Upate atomic ops in nacl_io

Remove unneeded lock in mount_dev, mount root nodes do the locking.
Add atomic and, or, xor.
Remove lock from RefObject.
Create SimpleLock.

BUG=259563
R=binji@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=211612

[noelallen1, 2013-07-11 23:54:46]

ptal

[binji, 2013-07-12 00:51:23]

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
File native_client_sdk/src/libraries/sdk_util/atomicops.h (right):

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/atomicops.h:24: inline Atomic32
AtomicAndFetch(volatile Atomic32* ptr, Atomic32 value) {
these are not used, maybe add them in a CL that uses them?

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
File native_client_sdk/src/libraries/sdk_util/simple_lock.h (right):

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:12: class AutoLock;
not necessary

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:20: class SimpleLock {
Why SimpleLock? Maybe just Lock?

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:30: operator
pthread_mutex_t*() const { return &lock_; }
We should just have a function for this, rather than an implicit conversion.

[noelallen1, 2013-07-12 23:18:47]

Converted AutoLock to take SimpleLock references.
Added AUTO_LOCK macro so allow us to plug in debugging.

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
File native_client_sdk/src/libraries/sdk_util/atomicops.h (right):

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/atomicops.h:24: inline Atomic32
AtomicAndFetch(volatile Atomic32* ptr, Atomic32 value) {
On 2013/07/12 00:51:23, binji wrote:
> these are not used, maybe add them in a CL that uses them?

I though it made more sense to break them out with the locking than to make the
epoll change more complex.

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
File native_client_sdk/src/libraries/sdk_util/simple_lock.h (right):

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:12: class AutoLock;
On 2013/07/12 00:51:23, binji wrote:
> not necessary

Done.

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:20: class SimpleLock {
On 2013/07/12 00:51:23, binji wrote:
> Why SimpleLock? Maybe just Lock? 

Less likely to collide with a name.

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:30: operator
pthread_mutex_t*() const { return &lock_; }
On 2013/07/12 00:51:23, binji wrote:
> We should just have a function for this, rather than an implicit conversion.

Done.

[binji, 2013-07-13 00:16:47]

lgtm

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
File native_client_sdk/src/libraries/sdk_util/simple_lock.h (right):

https://codereview.chromium.org/18644009/diff/1/native_client_sdk/src/librari...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:20: class SimpleLock {
On 2013/07/12 23:18:47, noelallen1 wrote:
> On 2013/07/12 00:51:23, binji wrote:
> > Why SimpleLock? Maybe just Lock? 
> 
> Less likely to collide with a name.

OK, can we change it to Lock after I add namespaces?

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
File native_client_sdk/src/libraries/nacl_io/kernel_handle.cc (right):

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
native_client_sdk/src/libraries/nacl_io/kernel_handle.cc:81: offs_ += *cnt;
this is slightly different behavior from before: this used to check for *cnt
being negative. Not worth keeping?

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
File native_client_sdk/src/libraries/nacl_io/kernel_handle.h (right):

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
native_client_sdk/src/libraries/nacl_io/kernel_handle.h:33: // Distatches Read,
Write, GetDents to atomically update offs_.
sp: Dispatches?

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
File native_client_sdk/src/libraries/sdk_util/atomicops.h (right):

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
native_client_sdk/src/libraries/sdk_util/atomicops.h:63: return
InterlockedOr(ptr, value);
xor?

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
File native_client_sdk/src/libraries/sdk_util/auto_lock.h (right):

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
native_client_sdk/src/libraries/sdk_util/auto_lock.h:15: // debugging perposes.
sp: purposes

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
File native_client_sdk/src/libraries/sdk_util/simple_lock.h (right):

https://codereview.chromium.org/18644009/diff/13002/native_client_sdk/src/lib...
native_client_sdk/src/libraries/sdk_util/simple_lock.h:32: friend class
AutoLock;
doesn't need to be a friend now, does it?

[noelallen1, 2013-07-14 23:48:12]

Committed patchset #7 manually as r211612 (presubmit successful).

[nfullagar1, 2013-07-18 00:18:06]

https://codereview.chromium.org/18644009/diff/30001/native_client_sdk/src/lib...
File native_client_sdk/src/libraries/sdk_util/thread_pool.cc (right):

https://codereview.chromium.org/18644009/diff/30001/native_client_sdk/src/lib...
native_client_sdk/src/libraries/sdk_util/thread_pool.cc:63: return
AtomicAddFetch(&counter_, 1);
This should return the decremented task counter, but is incrementing it instead.
 This change is breaking all examples that are using threadpool.