3.16.84-rc1 review patch. If anyone has any objections, please let me know. ------------------ From: Mathias Krause <minipli@xxxxxxxxxxxxxx> commit 350ef88e7e922354f82a931897ad4a4ce6c686ff upstream. If the algorithm we're parallelizing is asynchronous we might change CPUs between padata_do_parallel() and padata_do_serial(). However, we don't expect this to happen as we need to enqueue the padata object into the per-cpu reorder queue we took it from, i.e. the same-cpu's parallel queue. Ensure we're not switching CPUs for a given padata object by tracking the CPU within the padata object. If the serial callback gets called on the wrong CPU, defer invoking padata_reorder() via a kernel worker on the CPU we're expected to run on. Signed-off-by: Mathias Krause <minipli@xxxxxxxxxxxxxx> Signed-off-by: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx> --- include/linux/padata.h | 2 ++ kernel/padata.c | 20 +++++++++++++++++++- 2 files changed, 21 insertions(+), 1 deletion(-) --- a/include/linux/padata.h +++ b/include/linux/padata.h @@ -37,6 +37,7 @@ * @list: List entry, to attach to the padata lists. * @pd: Pointer to the internal control structure. * @cb_cpu: Callback cpu for serializatioon. + * @cpu: Cpu for parallelization. * @seq_nr: Sequence number of the parallelized data object. * @info: Used to pass information from the parallel to the serial function. * @parallel: Parallel execution function. @@ -46,6 +47,7 @@ struct padata_priv { struct list_head list; struct parallel_data *pd; int cb_cpu; + int cpu; int info; void (*parallel)(struct padata_priv *padata); void (*serial)(struct padata_priv *padata); --- a/kernel/padata.c +++ b/kernel/padata.c @@ -130,6 +130,7 @@ int padata_do_parallel(struct padata_ins padata->cb_cpu = cb_cpu; target_cpu = padata_cpu_hash(pd); + padata->cpu = target_cpu; queue = per_cpu_ptr(pd->pqueue, target_cpu); spin_lock(&queue->parallel.lock); @@ -367,10 +368,21 @@ void padata_do_serial(struct padata_priv int cpu; struct padata_parallel_queue *pqueue; struct parallel_data *pd; + int reorder_via_wq = 0; pd = padata->pd; cpu = get_cpu(); + + /* We need to run on the same CPU padata_do_parallel(.., padata, ..) + * was called on -- or, at least, enqueue the padata object into the + * correct per-cpu queue. + */ + if (cpu != padata->cpu) { + reorder_via_wq = 1; + cpu = padata->cpu; + } + pqueue = per_cpu_ptr(pd->pqueue, cpu); spin_lock(&pqueue->reorder.lock); @@ -387,7 +399,13 @@ void padata_do_serial(struct padata_priv put_cpu(); - padata_reorder(pd); + /* If we're running on the wrong CPU, call padata_reorder() via a + * kernel worker. + */ + if (reorder_via_wq) + queue_work_on(cpu, pd->pinst->wq, &pqueue->reorder_work); + else + padata_reorder(pd); } EXPORT_SYMBOL(padata_do_serial);