[Cbe-oss-dev] [PATCH] spufs: avoid using backing operation for wbox write

André Detsch adetsch at br.ibm.com
Fri Aug 10 03:21:19 EST 2007


Arnd Bergmann wrote:
> Well, 21000002 means that the program has successfully exited (exit code 0),
> which the expected behaviour. So what's the problem now?

The context is actually not finished. Simulator tells me that we are 
still at rdch operation, and ch29 stalled.

Follows some information which might be useful:

# Trace of ctx=0xc00000000fda5300: some relevant steps before
# waiting forever for a mb write, which was somehow lost
spufs_run_spu
     spu_reacquire_runnable
         spu_acquire_runnable
             spu_activate
                 spu_bind_context
                     spu_restore
                         restore_csa
                             restore_spu_mb
                             # spu_status_R = 3ffc0002
                             restore_spu_runcntl
                             # spu_status_R = 1
                             # spu_status_R = 21000002
         # spu_status_R = 21000002
         spu_run_init
             ctx->ops->npc_write(ctx, *npc);
             # spu_status_R = 21000002
             out_be64(&ctx->spu->priv2->spu_privcntl_RW, mode);
             # spu_status_R = 21000002
             ctx->ops->runcntl_write(ctx, SPU_RUNCNTL_RUNNABLE);
             # spu_status_R = 9
     # we get back to spufs_run_spu, never going beyond
     # spufs_wait(ctx->stop_wq, spu_stopped(ctx, &status)),
     # since spu_status_R is kept as '9'

--
Andre Detsch



More information about the cbe-oss-dev mailing list