Linaro-mm-sig

linaro-mm-sig@lists.linaro.org

16 participants
2940 discussions

Re: [RFC PATCH v3 10/12] tcp: RX path for devmem TCP

by Jakub Kicinski

On Sun, 5 Nov 2023 18:44:09 -0800 Mina Almasry wrote: > + if (!skb_frags_not_readable(skb)) { nit: maybe call the helper skb_frags_readable() and flip the polarity, avoid double negatives.

1 year, 8 months

Re: [RFC PATCH v3 02/12] net: page_pool: create hooks for custom page providers

by Jakub Kicinski

On Sun, 5 Nov 2023 18:44:01 -0800 Mina Almasry wrote: > diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h > index 6fc5134095ed..d4bea053bb7e 100644 > --- a/include/net/page_pool/types.h > +++ b/include/net/page_pool/types.h > @@ -60,6 +60,8 @@ struct page_pool_params { > int nid; > struct device *dev; > struct napi_struct *napi; > + u8 memory_provider; > + void *mp_priv; > enum dma_data_direction dma_dir; > unsigned int max_len; > unsigned int offset; you should rebase on top of net-next More importantly I was expecting those fields to be gone from params. The fact that the page pool is configured to a specific provider should be fully transparent to the driver, driver should just tell the core what queue its creating the pool from and if there's a dmabuf bound for that queue - out pops a pp backed by the dmabuf. My RFC had the page pool params fields here as a hack.

1 year, 8 months

Re: [RFC PATCH v3 12/12] selftests: add ncdevmem, netcat for devmem TCP

by Jakub Kicinski

On Sun, 5 Nov 2023 18:44:11 -0800 Mina Almasry wrote: > + if (ynl_subscribe(*ys, "mgmt")) > + goto err_close; Why? :)

1 year, 8 months

Re: [RFC PATCH v3 09/12] net: add support for skbs with unreadable frags

by Jakub Kicinski

On Tue, 7 Nov 2023 14:23:20 -0800 Stanislav Fomichev wrote: > Can we mark a socket as devmem-only? Do we have any use-case for those > hybrid setups? Or, let me put it that way: do we expect API callers > to handle both linear and non-linear cases correctly? > As a consumer of the previous versions of these apis internally, > I find all those corner cases confusing :-( Hence trying to understand > whether we can make it a bit more rigid and properly defined upstream. FWIW I'd also prefer to allow mixing. "Some NICs" can decide HDS very flexibly, incl. landing full jumbo frames into the "headers". There's no sender API today to signal how to mark the data for selective landing, but if Mina already has the rx side written to allow that...

1 year, 8 months

Re: [RFC PATCH v3 03/12] net: netdev netlink api to bind dma-buf to a net device

by Jakub Kicinski

On Sun, 5 Nov 2023 18:44:02 -0800 Mina Almasry wrote: > + - > + name: queues > + doc: receive queues to bind the dma-buf to. > + type: u32 > + multi-attr: true I think that you should throw in the queue type. I know you made the op imply RX: > + - > + name: bind-rx but if we decide to create a separate "type" for some magic queue type one day we'll have to ponder how to extend this API IMHO queue should be identified by a <type, id> tuple, always.

1 year, 8 months

Re: [RFC PATCH v3 06/12] memory-provider: dmabuf devmem memory provider

by Jakub Kicinski

On Sun, 5 Nov 2023 18:44:05 -0800 Mina Almasry wrote: > +static int mp_dmabuf_devmem_init(struct page_pool *pool) > +{ > + struct netdev_dmabuf_binding *binding = pool->mp_priv; > + > + if (!binding) > + return -EINVAL; > + > + if (pool->p.flags & PP_FLAG_DMA_MAP || > + pool->p.flags & PP_FLAG_DMA_SYNC_DEV) > + return -EOPNOTSUPP; This looks backwards, we should _force_ the driver to use the dma mapping built into the page pool APIs, to isolate the driver from how the DMA addr actually gets obtained. Right? Maybe seeing driver patches would illuminate.

1 year, 8 months

Re: [RFC PATCH v3 10/12] tcp: RX path for devmem TCP

by Jakub Kicinski

On Thu, 09 Nov 2023 12:05:37 +0100 Paolo Abeni wrote: > > I suppose we just disagree on the elegance of the API. > > FWIW, I think sockopt +cmsg is the right API. FWIW it's fine by me as well.

1 year, 8 months

Re: [RFC PATCH v3 12/12] selftests: add ncdevmem, netcat for devmem TCP

by Jakub Kicinski

My brain is slightly fried after trying to catch up on the thread for close to 2h. So forgive me if I'm missing something. This applies to all emails I'm about to send :) On Sun, 5 Nov 2023 18:44:11 -0800 Mina Almasry wrote: > + trigger_device_reset(); The user space must not be responsible for the reset. We can add some temporary "recreate page pools" ndo until the queue API is ready. But it should not be visible to the user in any way. And then the kernel can issue the same reset when the netlink socket dies to flush device free lists. Maybe we should also add a "allow device/all-queues reload" flag to the netlink API to differentiate drivers which can't implement full queue API later on. We want to make sure the defaults work well in our "target design", rather than at the first stage. And target design will reload queues one by one.

1 year, 8 months

Re: [RFC PATCH v2] drm/test: add a test suite for GEM objects backed by shmem

by Maxime Ripard

On Wed, Nov 08, 2023 at 02:42:03PM +0100, Marco Pagani wrote: > This patch introduces an initial KUnit test suite for GEM objects > backed by shmem buffers. > > Suggested-by: Javier Martinez Canillas <javierm(a)redhat.com> > Signed-off-by: Marco Pagani <marpagan(a)redhat.com> > > v2: > - Improved description of test cases > - Cleaner error handling using KUnit actions > - Alphabetical order in Kconfig and Makefile > --- > drivers/gpu/drm/Kconfig | 9 +- > drivers/gpu/drm/tests/Makefile | 5 +- > drivers/gpu/drm/tests/drm_gem_shmem_test.c | 381 +++++++++++++++++++++ > 3 files changed, 389 insertions(+), 6 deletions(-) > create mode 100644 drivers/gpu/drm/tests/drm_gem_shmem_test.c > > diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig > index 3eee8636f847..a2551c8c393a 100644 > --- a/drivers/gpu/drm/Kconfig > +++ b/drivers/gpu/drm/Kconfig > @@ -76,14 +76,15 @@ config DRM_KUNIT_TEST > tristate "KUnit tests for DRM" if !KUNIT_ALL_TESTS > depends on DRM && KUNIT > select PRIME_NUMBERS > + select DRM_BUDDY > select DRM_DISPLAY_DP_HELPER > select DRM_DISPLAY_HELPER > - select DRM_LIB_RANDOM > - select DRM_KMS_HELPER > - select DRM_BUDDY > + select DRM_EXEC > select DRM_EXPORT_FOR_TESTS if m > + select DRM_GEM_SHMEM_HELPER > + select DRM_KMS_HELPER > select DRM_KUNIT_TEST_HELPERS > - select DRM_EXEC > + select DRM_LIB_RANDOM > default KUNIT_ALL_TESTS > help > This builds unit tests for DRM. This option is not useful for > diff --git a/drivers/gpu/drm/tests/Makefile b/drivers/gpu/drm/tests/Makefile > index ba7baa622675..d6183b3d7688 100644 > --- a/drivers/gpu/drm/tests/Makefile > +++ b/drivers/gpu/drm/tests/Makefile > @@ -9,15 +9,16 @@ obj-$(CONFIG_DRM_KUNIT_TEST) += \ > drm_connector_test.o \ > drm_damage_helper_test.o \ > drm_dp_mst_helper_test.o \ > + drm_exec_test.o \ > drm_format_helper_test.o \ > drm_format_test.o \ > drm_framebuffer_test.o \ > + drm_gem_shmem_test.o \ > drm_managed_test.o \ > drm_mm_test.o \ > drm_modes_test.o \ > drm_plane_helper_test.o \ > drm_probe_helper_test.o \ > - drm_rect_test.o \ > - drm_exec_test.o > + drm_rect_test.o Thanks for reordering the tests and symbols, but they should part of a preliminary patch. > CFLAGS_drm_mm_test.o := $(DISABLE_STRUCTLEAK_PLUGIN) > diff --git a/drivers/gpu/drm/tests/drm_gem_shmem_test.c b/drivers/gpu/drm/tests/drm_gem_shmem_test.c > new file mode 100644 > index 000000000000..983380490673 > --- /dev/null > +++ b/drivers/gpu/drm/tests/drm_gem_shmem_test.c > @@ -0,0 +1,381 @@ > +// SPDX-License-Identifier: GPL-2.0 > +/* > + * KUnit test suite for GEM objects backed by shmem buffers > + * > + * Copyright (C) 2023 Red Hat, Inc. > + * > + * Author: Marco Pagani <marpagan(a)redhat.com> > + */ > + > +#include <linux/dma-buf.h> > +#include <linux/iosys-map.h> > +#include <linux/sizes.h> > + > +#include <kunit/test.h> > + > +#include <drm/drm_device.h> > +#include <drm/drm_drv.h> > +#include <drm/drm_gem.h> > +#include <drm/drm_gem_shmem_helper.h> > +#include <drm/drm_kunit_helpers.h> > + > +#define TEST_SIZE SZ_1M > +#define TEST_BYTE 0xae > + > +struct fake_dev { > + struct drm_device drm_dev; > + struct device *dev; AFAICS, you're not using the dev pointer anywhere. You could remove the fake_dev struct entirely and pass the drm_device pointer in test->priv > +}; > + > +/* > + * Wrappers to avoid an explicit type casting when passing action > + * functions to kunit_add_action(). > + */ > +static void kfree_wrapper(void *p) > +{ > + kfree(p); > +} > + > +static void sg_free_table_wrapper(void *sgt) > +{ > + sg_free_table(sgt); > +} > + > +static void drm_gem_shmem_free_wrapper(void *shmem) > +{ > + drm_gem_shmem_free(shmem); > +} I think you need to explicitly cast the pointer (or do a temporary assignment to the proper type) to avoid a compiler warning. > +/* > + * Test creating a shmem GEM object backed by shmem buffer. The test > + * case succeeds if the GEM object is successfully allocated with the > + * shmem file node and object functions attributes set, and the size > + * attribute is equal to the correct size. > + */ Thanks for all those comments, it's super helpful > +static void drm_gem_shmem_test_obj_create(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + KUNIT_EXPECT_EQ(test, shmem->base.size, TEST_SIZE); > + KUNIT_EXPECT_NOT_NULL(test, shmem->base.filp); > + KUNIT_EXPECT_NOT_NULL(test, shmem->base.funcs); > + > + drm_gem_shmem_free(shmem); > +} > + > +/* > + * Test creating a shmem GEM object from a scatter/gather table exported > + * via a DMA-BUF. The test case succeed if the GEM object is successfully > + * created with the shmem file node attribute equal to NULL and the sgt > + * attribute pointing to the scatter/gather table that has been imported. > + */ > +static void drm_gem_shmem_test_obj_create_private(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + struct drm_gem_object *gem_obj; > + struct dma_buf buf_mock; > + struct dma_buf_attachment attach_mock; > + struct sg_table *sgt; > + char *buf; > + int ret; > + > + /* Create a mock scatter/gather table */ > + buf = kunit_kzalloc(test, TEST_SIZE, GFP_KERNEL); > + KUNIT_ASSERT_NOT_NULL(test, buf); > + > + sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); > + KUNIT_ASSERT_NOT_NULL(test, sgt); > + > + ret = kunit_add_action_or_reset(test, kfree_wrapper, sgt); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = sg_alloc_table(sgt, 1, GFP_KERNEL); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = kunit_add_action_or_reset(test, sg_free_table_wrapper, sgt); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + sg_init_one(sgt->sgl, buf, TEST_SIZE); > + > + /* Init a mock DMA-BUF */ > + buf_mock.size = TEST_SIZE; > + attach_mock.dmabuf = &buf_mock; > + > + gem_obj = drm_gem_shmem_prime_import_sg_table(&fdev->drm_dev, &attach_mock, sgt); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, gem_obj); > + KUNIT_EXPECT_EQ(test, gem_obj->size, TEST_SIZE); > + KUNIT_EXPECT_NULL(test, gem_obj->filp); > + KUNIT_EXPECT_NOT_NULL(test, gem_obj->funcs); > + > + /* The scatter/gather table will be freed by drm_gem_shmem_free */ > + kunit_remove_action(test, sg_free_table_wrapper, sgt); > + kunit_remove_action(test, kfree_wrapper, sgt); > + > + shmem = to_drm_gem_shmem_obj(gem_obj); > + KUNIT_EXPECT_PTR_EQ(test, shmem->sgt, sgt); > + > + drm_gem_shmem_free(shmem); > +} > + > +/* > + * Test pinning backing pages for a shmem GEM object. The test case > + * succeeds if a suitable number of backing pages are allocated, and > + * the pages table counter attribute is increased by one. > + */ > +static void drm_gem_shmem_test_pin_pages(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + int i, ret; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + KUNIT_EXPECT_NULL(test, shmem->pages); > + KUNIT_EXPECT_EQ(test, shmem->pages_use_count, 0); > + > + ret = kunit_add_action_or_reset(test, drm_gem_shmem_free_wrapper, shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = drm_gem_shmem_pin(shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + KUNIT_ASSERT_NOT_NULL(test, shmem->pages); > + KUNIT_EXPECT_EQ(test, shmem->pages_use_count, 1); > + > + for (i = 0; i < (shmem->base.size >> PAGE_SHIFT); i++) > + KUNIT_ASSERT_NOT_NULL(test, shmem->pages[i]); > + > + drm_gem_shmem_unpin(shmem); > + KUNIT_EXPECT_NULL(test, shmem->pages); > + KUNIT_EXPECT_EQ(test, shmem->pages_use_count, 0); > +} > + > +/* > + * Test creating a virtual mapping for a shmem GEM object. The test > + * case succeeds if the backing memory is mapped and the reference > + * counter for virtual mapping is increased by one. Moreover, the test > + * case writes and then reads a test pattern over the mapped memory. > + */ > +static void drm_gem_shmem_test_vmap(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + struct iosys_map map; > + int ret, i; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + KUNIT_EXPECT_NULL(test, shmem->vaddr); > + KUNIT_EXPECT_EQ(test, shmem->vmap_use_count, 0); > + > + ret = kunit_add_action_or_reset(test, drm_gem_shmem_free_wrapper, shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = drm_gem_shmem_vmap(shmem, &map); > + KUNIT_ASSERT_EQ(test, ret, 0); > + KUNIT_ASSERT_NOT_NULL(test, shmem->vaddr); > + KUNIT_ASSERT_FALSE(test, iosys_map_is_null(&map)); > + KUNIT_EXPECT_EQ(test, shmem->vmap_use_count, 1); > + > + iosys_map_memset(&map, 0, TEST_BYTE, TEST_SIZE); > + for (i = 0; i < TEST_SIZE; i++) > + KUNIT_EXPECT_EQ(test, iosys_map_rd(&map, i, u8), TEST_BYTE); > + > + drm_gem_shmem_vunmap(shmem, &map); > + KUNIT_EXPECT_NULL(test, shmem->vaddr); > + KUNIT_EXPECT_EQ(test, shmem->vmap_use_count, 0); > +} > + > +/* > + * Test exporting a scatter/gather table of pinned pages suitable for > + * PRIME usage from a shmem GEM object. The test case succeeds if a > + * scatter/gather table large enough to accommodate the backing memory > + * is successfully exported. > + */ > +static void drm_gem_shmem_test_get_pages_sgt(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + struct sg_table *sgt; > + struct scatterlist *sg; > + unsigned int si, len = 0; > + int ret; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + > + ret = kunit_add_action_or_reset(test, drm_gem_shmem_free_wrapper, shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = drm_gem_shmem_pin(shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + sgt = drm_gem_shmem_get_sg_table(shmem); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, sgt); > + KUNIT_EXPECT_NULL(test, shmem->sgt); > + > + ret = kunit_add_action_or_reset(test, sg_free_table_wrapper, sgt); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + for_each_sgtable_sg(sgt, sg, si) { > + KUNIT_EXPECT_NOT_NULL(test, sg); > + len += sg->length; > + } > + > + KUNIT_EXPECT_GE(test, len, TEST_SIZE); > +} > + > +/* > + * Test pinning pages and exporting a scatter/gather table suitable for > + * driver usage from a shmem GEM object. The test case succeeds if the > + * backing pages are pinned and a scatter/gather table large enough to > + * accommodate the backing memory is successfully exported. > + */ > +static void drm_gem_shmem_test_get_sg_table(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + struct sg_table *sgt; > + struct scatterlist *sg; > + unsigned int si, len, ret = 0; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + > + ret = kunit_add_action_or_reset(test, drm_gem_shmem_free_wrapper, shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + /* The scatter/gather table will be freed by drm_gem_shmem_free */ > + sgt = drm_gem_shmem_get_pages_sgt(shmem); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, sgt); > + KUNIT_ASSERT_NOT_NULL(test, shmem->pages); > + KUNIT_EXPECT_EQ(test, shmem->pages_use_count, 1); > + KUNIT_EXPECT_PTR_EQ(test, sgt, shmem->sgt); > + > + for_each_sgtable_sg(sgt, sg, si) { > + KUNIT_EXPECT_NOT_NULL(test, sg); > + len += sg->length; > + } > + > + KUNIT_EXPECT_GE(test, len, TEST_SIZE); > +} > + > +/* > + * Test updating the madvise state of a shmem GEM object. The test > + * case checks that the function for setting madv updates it only if > + * its current value is greater or equal than zero and returns false > + * if it has a negative value. > + */ > +static void drm_gem_shmem_test_madvise(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + int ret; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + KUNIT_ASSERT_EQ(test, shmem->madv, 0); > + > + ret = kunit_add_action_or_reset(test, drm_gem_shmem_free_wrapper, shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = drm_gem_shmem_madvise(shmem, 1); > + KUNIT_EXPECT_TRUE(test, ret); > + KUNIT_ASSERT_EQ(test, shmem->madv, 1); > + > + /* Set madv to a negative value */ > + ret = drm_gem_shmem_madvise(shmem, -1); > + KUNIT_EXPECT_FALSE(test, ret); > + KUNIT_ASSERT_EQ(test, shmem->madv, -1); > + > + /* Check that madv cannot be set back to a positive value */ > + ret = drm_gem_shmem_madvise(shmem, 0); > + KUNIT_EXPECT_FALSE(test, ret); > + KUNIT_ASSERT_EQ(test, shmem->madv, -1); > +} > + > +/* > + * Test purging a shmem GEM object. First, assert that a newly created > + * shmem GEM object is not purgeable. Then, set madvise to a positive > + * value and call drm_gem_shmem_get_pages_sgt() to pin and dma-map the > + * backing pages. Finally, assert that the shmem GEM object is now > + * purgeable and purge it. > + */ > +static void drm_gem_shmem_test_purge(struct kunit *test) > +{ > + struct fake_dev *fdev = test->priv; > + struct drm_gem_shmem_object *shmem; > + struct sg_table *sgt; > + int ret; > + > + shmem = drm_gem_shmem_create(&fdev->drm_dev, TEST_SIZE); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, shmem); > + > + ret = kunit_add_action_or_reset(test, drm_gem_shmem_free_wrapper, shmem); > + KUNIT_ASSERT_EQ(test, ret, 0); > + > + ret = drm_gem_shmem_is_purgeable(shmem); > + KUNIT_EXPECT_FALSE(test, ret); > + > + ret = drm_gem_shmem_madvise(shmem, 1); > + KUNIT_EXPECT_TRUE(test, ret); > + > + /* The scatter/gather table will be freed by drm_gem_shmem_free */ > + sgt = drm_gem_shmem_get_pages_sgt(shmem); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, sgt); > + > + ret = drm_gem_shmem_is_purgeable(shmem); > + KUNIT_EXPECT_TRUE(test, ret); > + > + drm_gem_shmem_purge(shmem); > + KUNIT_EXPECT_NULL(test, shmem->pages); > + KUNIT_EXPECT_NULL(test, shmem->sgt); > + KUNIT_EXPECT_EQ(test, shmem->madv, -1); > +} > + > +static int drm_gem_shmem_test_init(struct kunit *test) > +{ > + struct fake_dev *fdev; > + struct device *dev; > + > + /* Allocate a parent device */ > + dev = drm_kunit_helper_alloc_device(test); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, dev); > + > + /* > + * The DRM core will automatically initialize the GEM core and create > + * a DRM Memory Manager object which provides an address space pool > + * for GEM objects allocation. > + */ > + fdev = drm_kunit_helper_alloc_drm_device(test, dev, struct fake_dev, > + drm_dev, DRIVER_GEM); > + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, fdev); > + > + fdev->dev = dev; > + test->priv = fdev; > + > + return 0; > +} > + > +static struct kunit_case drm_gem_shmem_test_cases[] = { > + KUNIT_CASE(drm_gem_shmem_test_obj_create), > + KUNIT_CASE(drm_gem_shmem_test_obj_create_private), > + KUNIT_CASE(drm_gem_shmem_test_pin_pages), > + KUNIT_CASE(drm_gem_shmem_test_vmap), > + KUNIT_CASE(drm_gem_shmem_test_get_pages_sgt), > + KUNIT_CASE(drm_gem_shmem_test_get_sg_table), > + KUNIT_CASE(drm_gem_shmem_test_madvise), > + KUNIT_CASE(drm_gem_shmem_test_purge), > + {} > +}; > + > +static struct kunit_suite drm_gem_shmem_suite = { > + .name = "drm_gem_shmem", > + .init = drm_gem_shmem_test_init, > + .test_cases = drm_gem_shmem_test_cases > +}; > + > +kunit_test_suite(drm_gem_shmem_suite); Looks great otherwise, thanks! Maxime

1 year, 8 months

Re: [RFC PATCH v3 07/12] page-pool: device memory support

by Christian König

Am 09.11.23 um 04:20 schrieb Mina Almasry: > [SNIP] >> But we might be able to do something as folio is doing now, mm subsystem >> is still seeing 'struct folio/page', but other subsystem like slab is using >> 'struct slab', and there is still some common fields shared between >> 'struct folio' and 'struct slab'. >> > In my eyes this is almost exactly what I suggested in RFC v1 and got > immediately nacked with no room to negotiate. What we did for v1 is to > allocate struct pages for dma-buf to make dma-bufs look like struct > page to mm subsystem. Almost exactly what you're describing above. > It's a no-go. I don't think renaming struct page to netmem is going to > move the needle (it also re-introduces code-churn). What I feel like I > learnt is that dma-bufs are not struct pages and can't be made to look > like one, I think. Yeah, that pretty much hits the nail on the head. What was not mentioned yet and you could potentially try to do is to go the other way around. In other words instead of importing a DMA-buf file descriptor into the page-pool, you take some netdev page-pool pages and export them as DMA-buf handle. All you then need to do is to implement the necessary DMA-buf interface, e.g. wait for DMA-fences before accessing stuff, when you have an async operation install a DMA-fence so that other can wait for your operation to finish etc.. etc.. This still doesn't gives you peer 2 peer over for example the PCIe bus, but it would give you zero copy in the sense that a GPU or other acceleration device directly writes stuff into memory a network device can work with. We already have some similar functionality for at least Intel and AMD hw where an userptr mechanism is used to make malloced memory (backed by normal struct pages) available to the GPU. The problem with this approach is that most GPUs currently can't do good recoverable page faults which in turn makes the whole MMU notifier based approach just horrible inefficient. Just giving a few more pointers you might want to look into... Cheers, Christian.

1 year, 8 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

Linaro-mm-sig