Changeset 235d31d in mainline for kernel/generic/include/atomic.h

Timestamp:

2014-12-22T17:47:40Z (9 years ago)

Author:

Jakub Jermar <jakub@…>

Branches:

lfn, master, serial, ticket/834-toolchain-update, topic/msim-upgrade, topic/simplify-dev-export

Children:

8c7d5ad

Parents:

eae91e0 (diff), 759ea0d (diff)
Note: this is a merge changeset, the changes displayed below correspond to the merge itself.
Use the (diff) links above to see all the changes relative to each parent.

Message:

Merge the CHT pre-integration branch

This branch contains:

• the merge of lp:~adam-hraska+lp/helenos/rcu, which brings:

• a new preemptible kernel RCU variant called A-RCU,
• a preemptible variant of Podzimek's non-preemptible kernel RCU and
• a new variant of usersace RCU,
• a new concurrent hash table (CHT) implementation based on RCU,
• a deployment of CHT in kernel futex handling,
• a deployment of the userspace RCU in the implementation of upgradable futexes,

all described in Adam Hraska's master thesis named Read-Copy-Update
for HelenOS, defended in 2013 at MFF UK; furthemore, the branch
fixes two synchronization bugs in condvars and waitq, respectively:

• revid:adam.hraska+hos@gmail.com-20121116144921-3to9u1tn1sg07rg7
• revid:adam.hraska+hos@gmail.com-20121116173623-km7gwtqixwudpe66

• build fixes required to pass make check
• overhaul of ia64 and sparc64 trap handling, to allow exc_dispatch() to be used now when the kernel is more picky about CPU state accounting
• an important fix of the sparc64/sun4v preemptible trap handler
• various other fixes of issues discovered on non-x86 architectures

File:

• kernel/generic/include/atomic.h (modified) (1 diff)

kernel/generic/include/atomic.h

@@ -53 +53 @@
 }
 
+
+/*
+ * If the architecture does not provide operations that are atomic
+ * only with respect to the local cpu (eg exception handlers) and
+ * not other cpus, implement these cpu local atomic operations with
+ * full blown smp-safe atomics.
+ */
+#ifndef local_atomic_exchange
+#define local_atomic_exchange(var_addr, new_val) \
+        __atomic_exchange_n((var_addr), (new_val), __ATOMIC_RELAXED)
+#endif
+
+
+
 #endif