All of lore.kernel.org
 help / color / mirror / Atom feed
From: Lucas Stach <l.stach@pengutronix.de>
To: Russell King <linux@arm.linux.org.uk>,
	Christian Gmeiner <christian.gmeiner@gmail.com>
Cc: dri-devel@lists.freedesktop.org
Subject: [PATCH 37/48] staging: etnaviv: avoid pinning pages in CMA
Date: Fri, 25 Sep 2015 13:57:49 +0200	[thread overview]
Message-ID: <1443182280-15868-38-git-send-email-l.stach@pengutronix.de> (raw)
In-Reply-To: <1443182280-15868-1-git-send-email-l.stach@pengutronix.de>

From: Russell King <rmk+kernel@arm.linux.org.uk>

The CMA memory area is shared between CMA and other users.  In order to
allow CMA to work, non-CMA allocations in this area must be for movable
pages, so that a CMA allocation can reclaim its memory from other users
when required.

Page cache backed allocations, such as pages allocated via shmem, are by
default marked as "movable" allocations, which means that they can come
from the CMA region.

However, in etnaviv's case, we allocate shmem pages, and then we pin
these pages for the life of the buffer.  This prevents the pages being
reclaimed, and causes CMA failures.

new_inode() in fs/inode.c even says:

 * Allocates a new inode for given superblock. The default gfp_mask
 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
 * If HIGHMEM pages are unsuitable or it is known that pages allocated
 * for the page cache are not reclaimable or migratable,
 * mapping_set_gfp_mask() must be called with suitable flags on the
 * newly created inode's mapping

Etnaviv shmem pages are not reclaimable nor migratable once they're
pinned, so etnaviv is not conforming with this requirement.  Change the
gfp_mask to GFP_HIGHUSER, so we can still allocate from highmem, but not
from the CMA area.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 drivers/staging/etnaviv/etnaviv_gem.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/staging/etnaviv/etnaviv_gem.c b/drivers/staging/etnaviv/etnaviv_gem.c
index 2518f897fdd8..5c2462d6130d 100644
--- a/drivers/staging/etnaviv/etnaviv_gem.c
+++ b/drivers/staging/etnaviv/etnaviv_gem.c
@@ -640,6 +640,19 @@ static struct drm_gem_object *__etnaviv_gem_new(struct drm_device *dev,
 	} else {
 		to_etnaviv_bo(obj)->ops = &etnaviv_gem_shmem_ops;
 		ret = drm_gem_object_init(dev, obj, size);
+		if (ret == 0) {
+			struct address_space *mapping;
+
+			/*
+			 * Our buffers are kept pinned, so allocating them
+			 * from the MOVABLE zone is a really bad idea, and
+			 * conflicts with CMA.  See coments above new_inode()
+			 * why this is required _and_ expected if you're
+			 * going to pin these pages.
+			 */
+			mapping = file_inode(obj->filp)->i_mapping;
+			mapping_set_gfp_mask(mapping, GFP_HIGHUSER);
+		}
 	}
 
 	if (ret)
-- 
2.5.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

  parent reply	other threads:[~2015-09-25 11:58 UTC|newest]

Thread overview: 100+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-09-11 14:10 [PATCH RFCv2 0/4] Etnaviv DRM driver again Lucas Stach
2015-09-11 14:10 ` [PATCH RFCv2 1/4] of: Add vendor prefix for Vivante Corporation Lucas Stach
2015-09-11 14:10 ` [PATCH RFCv2 2/4] staging: etnaviv: add devicetree bindings Lucas Stach
2015-09-11 14:10 ` [PATCH RFCv2 3/4] staging: etnaviv: add drm driver Lucas Stach
2015-09-14 13:16   ` Rob Clark
2015-09-16  7:56     ` Russell King - ARM Linux
2015-09-16  6:11   ` Christian Gmeiner
2015-09-16  7:49     ` Russell King - ARM Linux
2015-09-16 15:30     ` Lucas Stach
2015-09-16  8:04   ` Russell King - ARM Linux
2015-09-16 10:42     ` Christian Gmeiner
2015-09-16 15:36     ` Lucas Stach
2015-09-25 11:57     ` [PATCH 00/48] Etnaviv changes RFCv1->RFCv2 Lucas Stach
2015-09-25 11:57       ` [PATCH 01/48] staging: etnaviv: avoid holding struct_mutex over dma_alloc_coherent() Lucas Stach
2015-09-25 11:57       ` [PATCH 02/48] staging: etnaviv: restructure iommu handling Lucas Stach
2015-09-25 11:57       ` [PATCH 03/48] staging: etnaviv: remove compat MMU code Lucas Stach
2015-09-25 12:18         ` Russell King - ARM Linux
2015-10-21 11:35           ` Russell King - ARM Linux
2015-10-21 12:37             ` Lucas Stach
2015-10-21 13:37               ` Russell King - ARM Linux
2015-10-21 14:53                 ` Lucas Stach
2015-10-21 15:13                   ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 04/48] staging: etnaviv: clean up public API (part 2) Lucas Stach
2015-09-25 11:57       ` [PATCH 05/48] staging: etnaviv: rename last remaining msm_* symbols Lucas Stach
2015-09-25 11:57       ` [PATCH 06/48] staging: etnaviv: rename last remaining bits from msm to etnaviv Lucas Stach
2015-09-25 11:57       ` [PATCH 07/48] staging: etnaviv: quiten down kernel log output Lucas Stach
2015-09-25 11:57       ` [PATCH 08/48] staging: etnaviv: add proper license header to all files Lucas Stach
2015-09-25 11:57       ` [PATCH 09/48] staging: etnaviv: add Dove GPU subsystem compatible Lucas Stach
2015-09-25 11:57       ` [PATCH 10/48] staging: etnaviv: fix missing error cleanups in etnaviv_load() Lucas Stach
2015-09-25 11:57       ` [PATCH 11/48] staging: etnaviv: fix off-by-one for iommu aperture end Lucas Stach
2015-09-25 11:57       ` [PATCH 12/48] staging: etnaviv: avoid lockdep circular dependency warning Lucas Stach
2015-09-25 12:20         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 13/48] staging: etnaviv: fix gpu debugfs show implementation Lucas Stach
2015-09-25 11:57       ` [PATCH 14/48] staging: etnaviv: use vm_insert_page() rather than vm_insert_mixed() Lucas Stach
2015-09-25 11:57       ` [PATCH 15/48] staging: etnaviv: etnaviv_gem_fault: reduce struct_mutex exposure Lucas Stach
2015-09-25 11:57       ` [PATCH 16/48] staging: etnaviv: give etnaviv_gem_mmap_offset() a sane behaviour Lucas Stach
2015-09-25 11:57       ` [PATCH 17/48] staging: etnaviv: allow etnaviv_ioctl_gem_info() locking to be interruptible Lucas Stach
2015-09-25 11:57       ` [PATCH 18/48] staging: etnaviv: make context a per-GPU thing Lucas Stach
2015-09-25 11:57       ` [PATCH 19/48] staging: etnaviv: switch to per-GPU fence completion implementation Lucas Stach
2015-09-25 11:57       ` [PATCH 20/48] staging: etnaviv: provide etnaviv_queue_work() Lucas Stach
2015-09-25 11:57       ` [PATCH 21/48] staging: etnaviv: use standard kernel types rather than stdint.h types Lucas Stach
2015-09-25 11:57       ` [PATCH 22/48] staging: etnaviv: no need to initialise a list_head Lucas Stach
2015-09-25 11:57       ` [PATCH 23/48] staging: etnaviv: fix oops caused by scanning for free blocks Lucas Stach
2015-09-25 11:57       ` [PATCH 24/48] staging: etnaviv: clean up etnaviv_iommu_unmap_gem() signature Lucas Stach
2015-09-25 11:57       ` [PATCH 25/48] staging: etnaviv: increase page table size to maximum Lucas Stach
2015-09-25 11:57       ` [PATCH 26/48] staging: etnaviv: fix BUG_ON when removing module Lucas Stach
2015-09-25 11:57       ` [PATCH 27/48] staging: etnaviv: provide a helper to load the GPU clock field Lucas Stach
2015-09-25 11:57       ` [PATCH 28/48] staging: etnaviv: rename GPU clock functions Lucas Stach
2015-09-25 11:57       ` [PATCH 29/48] staging: etnaviv: fix runtime resume Lucas Stach
2015-09-25 11:57       ` [PATCH 30/48] staging: etnaviv: drop event ring buffer tracking Lucas Stach
2015-09-25 11:57       ` [PATCH 31/48] staging: etnaviv: improve efficiency of command parser Lucas Stach
2015-09-25 11:57       ` [PATCH 32/48] staging: etnaviv: no point looking up the mapping for cmdstream bos Lucas Stach
2015-09-25 11:57       ` [PATCH 33/48] staging: etnaviv: copy submit command and bos in one go Lucas Stach
2015-09-25 11:57       ` [PATCH 34/48] staging: etnaviv: remove cmd buffer offset validation in submit_reloc() Lucas Stach
2015-09-25 11:57       ` [PATCH 35/48] staging: etnaviv: move mapping teardown into etnaviv_gem_free_object() Lucas Stach
2015-09-25 11:57       ` [PATCH 36/48] staging: etnaviv: add support for GEM_WAIT ioctl Lucas Stach
2015-09-25 11:57       ` Lucas Stach [this message]
2015-09-25 11:57       ` [PATCH 38/48] staging: etnaviv: fix 'ret' may be used uninitialized in this function Lucas Stach
2015-09-25 11:57       ` [PATCH 39/48] staging: etnaviv: fix error: 'etnaviv_gpu_hw_resume' defined but not used Lucas Stach
2015-09-25 11:57       ` [PATCH 40/48] staging: etnaviv: debugfs: add possibility to dump kernel buffer Lucas Stach
2015-10-21 11:38         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 41/48] staging: etnaviv: change etnaviv_buffer_init() to return prefetch Lucas Stach
2015-10-21 11:38         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 42/48] staging: etnaviv: implement simple hang recovery Lucas Stach
2015-10-21 15:43         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 43/48] staging: etnaviv: map all buffers to the GPU Lucas Stach
2015-10-21 15:23         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 44/48] staging: etnaviv: implement cache maintenance on cpu_(prep|fini) Lucas Stach
2015-10-21 15:23         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 45/48] staging: etnaviv: remove submit type Lucas Stach
2015-10-21 14:41         ` Russell King - ARM Linux
2015-09-25 11:57       ` [PATCH 46/48] staging: etnaviv: rewrite submit interface to use copy from user Lucas Stach
2015-10-21 14:41         ` Russell King - ARM Linux
2015-10-26 20:48         ` Russell King - ARM Linux
2015-10-27 10:46           ` Lucas Stach
2015-09-25 11:57       ` [PATCH 47/48] staging: etnaviv: don't use GEM buffer for internal ring buffer Lucas Stach
2015-10-21 14:51         ` Russell King - ARM Linux
2015-09-25 11:58       ` [PATCH 48/48] staging: etnaviv: remove CMDSTREAM GEM allocation from UAPI Lucas Stach
2015-10-21 15:29         ` Russell King - ARM Linux
2015-09-28  9:46       ` [PATCH 00/48] Etnaviv changes RFCv1->RFCv2 Christian Gmeiner
2015-09-28 10:39         ` Lucas Stach
2015-09-30  7:53           ` Christian Gmeiner
2015-10-01  8:50             ` Lucas Stach
2015-10-13  8:25             ` Lucas Stach
2015-10-20  7:20               ` Christian Gmeiner
2015-10-20  9:00                 ` Lucas Stach
2015-10-20  9:09                   ` Jon Nettleton
2015-10-20  9:40                     ` Christian Gmeiner
2015-10-20 10:42                     ` Fabio Estevam
2015-10-20  9:39                   ` Christian Gmeiner
2015-09-16 15:05   ` [PATCH RFCv2 3/4] staging: etnaviv: add drm driver Eric Anholt
2015-09-16 16:51     ` Russell King - ARM Linux
2015-09-16 18:43       ` Eric Anholt
2015-09-11 14:10 ` [PATCH RFCv2 4/4] ARM: imx6: add Vivante GPU nodes Lucas Stach
2015-09-11 14:15 ` [PATCH RFCv2 0/4] Etnaviv DRM driver again Lucas Stach
2015-10-20  9:36 ` Daniel Vetter
2015-10-21 17:04   ` Russell King - ARM Linux
2015-10-22  7:12     ` Daniel Vetter
2015-10-22  8:19       ` Lucas Stach
2015-10-22  8:42       ` Lucas Stach

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1443182280-15868-38-git-send-email-l.stach@pengutronix.de \
    --to=l.stach@pengutronix.de \
    --cc=christian.gmeiner@gmail.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=linux@arm.linux.org.uk \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.