[Cbe-oss-dev] [PATCH 4/4] spufs: fix timing dependent false return from spufs_run_spu

Jeremy Kerr jk at ozlabs.org
Fri Feb 8 15:50:41 EST 2008


From: Luke Browning <lukebr at linux.vnet.ibm.com>

Stop bits are only valid when the running bit is not set. Status bits carry
over from one invocation of spufs_run_spu() to another, so the RUNNING bit
gets added to the previous state of the register which may have been a remote
library call.  In this case, it looks like another library routine should be
invoked, but the spe is actually running.

This fixes a problem with a testcase that exercises the scheduler.

Signed-off-by: Luke Browning <lukebrowning at us.ibm.com>
Signed-off-by: Jeremy Kerr <jk at ozlabs.org>

---
 arch/powerpc/platforms/cell/spufs/run.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/cell/spufs/run.c b/arch/powerpc/platforms/cell/spufs/run.c
index f401e51..fca22e1 100644
--- a/arch/powerpc/platforms/cell/spufs/run.c
+++ b/arch/powerpc/platforms/cell/spufs/run.c
@@ -53,7 +53,7 @@ int spu_stopped(struct spu_context *ctx, u32 *stat)
 
 	stopped = SPU_STATUS_INVALID_INSTR | SPU_STATUS_SINGLE_STEP |
 		SPU_STATUS_STOPPED_BY_HALT | SPU_STATUS_STOPPED_BY_STOP;
-	if (*stat & stopped)
+	if (!(*stat & SPU_STATUS_RUNNING) && (*stat & stopped))
 		return 1;
 
 	dsisr = ctx->csa.dsisr;



More information about the cbe-oss-dev mailing list