Return-Path: From: Arnd Bergmann To: Paul Mackerras Subject: [PATCH] spufs: always map local store non-guarded Date: Thu, 2 Nov 2006 13:46:48 +0100 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Message-Id: <200611021346.49473.arnd@arndb.de> Cc: linuxppc-dev@ozlabs.org, linux-kernel@vger.kernel.org, michaele@au1.ibm.com, cbe-oss-dev@ozlabs.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , When fixing spufs to map the 'mem' file backing store cacheable, I incorrectly set the physical mapping to use both cache-inhibited and guarded mapping, which resulted in a serious performance degradation. Accessing the real local store memory needs to be cache-inhibited, in order to maintain data consistency, but since it is actual RAM, there is no point in a guarded mapping. Debugged-by: Michael Ellerman Signed-off-by: Arnd Bergmann --- This fixes a regression in 2.6.19, please merge. Index: linux-2.6/arch/powerpc/platforms/cell/spufs/file.c =================================================================== --- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/file.c +++ linux-2.6/arch/powerpc/platforms/cell/spufs/file.c @@ -104,11 +104,11 @@ spufs_mem_mmap_nopage(struct vm_area_str if (ctx->state == SPU_STATE_SAVED) { vma->vm_page_prot = __pgprot(pgprot_val(vma->vm_page_prot) - & ~(_PAGE_NO_CACHE | _PAGE_GUARDED)); + & ~_PAGE_NO_CACHE); page = vmalloc_to_page(ctx->csa.lscsa->ls + offset); } else { vma->vm_page_prot = __pgprot(pgprot_val(vma->vm_page_prot) - | _PAGE_NO_CACHE | _PAGE_GUARDED); + | _PAGE_NO_CACHE); page = pfn_to_page((ctx->spu->local_store_phys + offset) >> PAGE_SHIFT); }