It's doing GFP_KERNEL on the IO submission path, with spinlocks held. Probably this means that the mempool is now redundant. Signed-off-by: Andrew Morton --- 25-akpm/drivers/scsi/qla2xxx/qla_os.c | 2 +- 1 files changed, 1 insertion(+), 1 deletion(-) diff -puN drivers/scsi/qla2xxx/qla_os.c~qla2xxx-allocation-mode-fix drivers/scsi/qla2xxx/qla_os.c --- 25/drivers/scsi/qla2xxx/qla_os.c~qla2xxx-allocation-mode-fix Mon Aug 9 15:37:56 2004 +++ 25-akpm/drivers/scsi/qla2xxx/qla_os.c Mon Aug 9 15:38:12 2004 @@ -3590,7 +3590,7 @@ qla2x00_get_new_sp(scsi_qla_host_t *ha) { srb_t *sp; - sp = mempool_alloc(ha->srb_mempool, GFP_KERNEL); + sp = mempool_alloc(ha->srb_mempool, GFP_ATOMIC); if (sp) atomic_set(&sp->ref_count, 1); return (sp); _