From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758681AbZFWLfv (ORCPT ); Tue, 23 Jun 2009 07:35:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757365AbZFWLfn (ORCPT ); Tue, 23 Jun 2009 07:35:43 -0400 Received: from casper.infradead.org ([85.118.1.10]:58112 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754822AbZFWLfm (ORCPT ); Tue, 23 Jun 2009 07:35:42 -0400 Subject: Re: [PATCH] radeon: use vmalloc instead of kmalloc From: Peter Zijlstra To: Jerome Glisse Cc: airlied@gmail.com, dri-devel@lists.sf.net, linux-kernel@vger.kernel.org In-Reply-To: <1245691561-15336-1-git-send-email-jglisse@redhat.com> References: <1245691561-15336-1-git-send-email-jglisse@redhat.com> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Tue, 23 Jun 2009 13:35:36 +0200 Message-Id: <1245756936.19816.1652.camel@twins> Mime-Version: 1.0 X-Mailer: Evolution 2.26.1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2009-06-22 at 19:26 +0200, Jerome Glisse wrote: > We don't need to allocated contiguous pages in cs codepath > so use vmalloc instead. Best would be to not require >PAGE_SIZE allocations at all of course. But barring that, it would be great to have something like: ptr = kmalloc(size, GFP_KERNEL | __GFP_NOWARN); if (!ptr) ptr = vmalloc(size); Also, how long do these allocations live? vmalloc space can be quite limited (i386) and it can fragment too. > Signed-off-by: Jerome Glisse > --- > drivers/gpu/drm/radeon/radeon_cs.c | 20 ++++++++++---------- > 1 files changed, 10 insertions(+), 10 deletions(-) > > diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c > index b843f9b..54a9740 100644 > --- a/drivers/gpu/drm/radeon/radeon_cs.c > +++ b/drivers/gpu/drm/radeon/radeon_cs.c > @@ -45,11 +45,11 @@ int radeon_cs_parser_relocs(struct radeon_cs_parser *p) > chunk = &p->chunks[p->chunk_relocs_idx]; > /* FIXME: we assume that each relocs use 4 dwords */ > p->nrelocs = chunk->length_dw / 4; > - p->relocs_ptr = kcalloc(p->nrelocs, sizeof(void *), GFP_KERNEL); > + p->relocs_ptr = vmalloc(p->nrelocs * sizeof(void *)); > if (p->relocs_ptr == NULL) { > return -ENOMEM; > } > - p->relocs = kcalloc(p->nrelocs, sizeof(struct radeon_cs_reloc), GFP_KERNEL); > + p->relocs = vmalloc(p->nrelocs * sizeof(struct radeon_cs_reloc)); > if (p->relocs == NULL) { > return -ENOMEM; > } > @@ -103,7 +103,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data) > p->idx = 0; > p->chunk_ib_idx = -1; > p->chunk_relocs_idx = -1; > - p->chunks_array = kcalloc(cs->num_chunks, sizeof(uint64_t), GFP_KERNEL); > + p->chunks_array = vmalloc(cs->num_chunks * sizeof(uint64_t)); > if (p->chunks_array == NULL) { > return -ENOMEM; > } > @@ -113,7 +113,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data) > return -EFAULT; > } > p->nchunks = cs->num_chunks; > - p->chunks = kcalloc(p->nchunks, sizeof(struct radeon_cs_chunk), GFP_KERNEL); > + p->chunks = vmalloc(p->nchunks * sizeof(struct radeon_cs_chunk)); > if (p->chunks == NULL) { > return -ENOMEM; > } > @@ -139,7 +139,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data) > > p->chunks[i].kdata = NULL; > size = p->chunks[i].length_dw * sizeof(uint32_t); > - p->chunks[i].kdata = kzalloc(size, GFP_KERNEL); > + p->chunks[i].kdata = vmalloc(size); > if (p->chunks[i].kdata == NULL) { > return -ENOMEM; > } > @@ -179,13 +179,13 @@ static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error) > mutex_unlock(&parser->rdev->ddev->struct_mutex); > } > } > - kfree(parser->relocs); > - kfree(parser->relocs_ptr); > + vfree(parser->relocs); > + vfree(parser->relocs_ptr); > for (i = 0; i < parser->nchunks; i++) { > - kfree(parser->chunks[i].kdata); > + vfree(parser->chunks[i].kdata); > } > - kfree(parser->chunks); > - kfree(parser->chunks_array); > + vfree(parser->chunks); > + vfree(parser->chunks_array); > radeon_ib_free(parser->rdev, &parser->ib); > } >