- Linaro-mm-sig - lists.linaro.org

Re: [Linaro-mm-sig] Memory eviction in ttm

by Thomas Hellström

Hi Maarten! Broadening the audience a bit.. On 9/14/12 9:12 AM, Maarten Lankhorst wrote: > Op 13-09-12 23:00, Thomas Hellstrom schreef: >> On 09/13/2012 07:13 PM, Maarten Lankhorst wrote: >>> Hey >>> >>> Op 13-09-12 18:41, Thomas Hellstrom schreef: >>>> On 09/13/2012 05:19 PM, Maarten Lankhorst wrote: >>>>> Hey, >>>>> >>>>> Op 12-09-12 15:28, Thomas Hellstrom schreef: >>>>>> On 09/12/2012 02:48 PM, Maarten Lankhorst wrote: >>>>>>> Hey Thomas, >>>>>>> >>>>>>> I'm playing around with moving reservations from ttm to global, but how ttm >>>>>>> ttm is handling reservations is getting in the way. The code wants to move >>>>>>> the bo from the lru lock at the same time a reservation is made, but that >>>>>>> seems to be slightly too strict. It would really help me if that guarantee >>>>>>> is removed. >>>>>> Hi, Maarten. >>>>>> >>>>>> Removing that restriction is not really possible at the moment. >>>>>> Also the memory accounting code depends on this, and may cause reservations >>>>>> in the most awkward places. Since these reservations don't have a ticket >>>>>> they may and will cause deadlocks. So in short the restriction is there >>>>>> to avoid deadlocks caused by ticketless reservations. >>>>> I have finished the lockdep annotations now which seems to catch almost >>>>> all abuse I threw at it, so I'm feeling slightly more confident about moving >>>>> the locking order and reservations around. >>>> Maarten, moving reservations in TTM out of the lru lock is incorrect as the code is >>>> written now. If we want to move it out we need something for ticketless reservations >>>> >>>> I've been thinking of having a global hash table of tickets with the task struct pointer as the key, >>>> but even then, we'd need to be able to handle EBUSY errors on every operation that might try to >>>> reserve a buffer. >>>> >>>> The fact that lockdep doesn't complain isn't enough. There *will* be deadlock use-cases when TTM is handed >>>> the right data-set. >>>> >>>> Isn't there a way that a subsystem can register a callback to be performed to remove stuff from LRU and >>>> to take a pre-reservation lock? >>> What if multiple subsystems need those? You will end up with a deadlock again. >>> >>> I think it would be easier to change the code in ttm_bo.c to not assume the first >>> item on the lru list is really the least recently used, and assume the first item >>> that can be reserved without blocking IS the least recently used instead. >> So what would happen then is that we'd spin on the first item on the LRU list, since >> when reserving we must release the LRU lock, and if reserving fails, we thus >> need to restart LRU traversal. Typically after a schedule(). That's bad. >> >> So let's take a step back and analyze why the LRU lock has become a problem. >> From what I can tell, it's because you want to use per-object lock when reserving instead of a >> global reservation lock (that TTM could use as the LRU lock). Is that correct? >> and in that case, in what situation do you envision such a global lock being contended >> to the extent that it hurts performance? >> >>>>> Lockdep WILL complain about trying to use multiple tickets, doing ticketed >>>>> and unticketed blocking reservations mixed, etc. >>>>> >>>>> I want to remove the global fence_lock and make it a per buffer lock, with some >>>>> lockdep annotations it's perfectly legal to grab obj->fence_lock and obj2->fence_lock >>>>> if you have a reservation, but it should complain loudly about trying to take 2 fence_locks >>>>> at the same time without a reservation. >>>> Yes, TTM was previously using per buffer fence locks, and that works fine from a deadlock perspective, but >>>> it hurts performance. Fencing 200 buffers in a command submission (google-earth for example) will mean >>>> 198 unnecessary locks, each discarding the processor pipelines. Locking is a *slow* operation, particularly >>>> on systems with many processors, and I don't think it's a good idea to change that back, without analyzing >>>> the performance impact. There are reasons people are writing stuff like RCU to avoid locking... >>> So why don't we simply use RCU for fence pointers and get rid of the fence locking? :D >>> danvet originally suggested it as a joke but if you think about it, it would make a lot of sense for this usecase. >> I thought of that before, but the problem is you'd still need a spinlock to change the buffer's fence pointer, >> even if reading it becomes quick. > Actually, I changed lockdep annotations a bit to distinguish between the > cases where ttm_bo_wait is called without reservation, and ttm_bo_wait > is called with, as far as I can see there are only 2 places that do it without, > at least if I converted my git tree properly.. > > http://cgit.freedesktop.org/~mlankhorst/linux/log/?h=v10-wip > > First one is nouveau_bo_vma_del, this can be fixed easily. > Second one is ttm_bo_cleanup_refs and ttm_bo_cleanup_refs_or_queue, > if reservation is done first before ttm_bo_wait, the fence_lock could be > dropped entirely by adding smb_mb() in reserve and unreserve, functionally > there would be no difference. So if you can verify my lockdep annotations are > correct in the most recent commit wrt what's using ttm_bo_wait without reservation > we could remove the fence_lock entirely. > > ~Maarten Being able to wait for buffer idle or get the fence pointer without reserving is a fundamental property of TTM. Reservation is a long-term lock. The fence lock is a very short term lock. If I were to choose, I'd rather accept per-object fence locks than removing this property, but see below. Likewise, to be able to guarantee that a reserved object is not on any LRU list is also an important property. Removing that property will, in addition to the spin wait we've already discussed make understanding TTM locking even more difficult, and I'd really like to avoid it. If this were a real performance problem we were trying to solve it would be easier to motivate changes in this area, but if it's just trying to avoid a global reservation lock and a global fence lock that will rarely if ever see any contention, I can't see the point. On the contrary, having per-object locks will be very costly when reserving / fencing many objects. As mentioned before, in the fence lock case it's been tried and removed, so I'd like to know the reasoning behind introducing it again, and in what situations you think the global locks will be contended. /Thomas

12 years, 8 months

3
2
0 0

[RFC][PATCH] fs/buffer.c: Revoke LRU when trying to drop buffers

by Laura Abbott

When a buffer is added to the LRU list, a reference is taken which is not dropped until the buffer is evicted from the LRU list. This is the correct behavior, however this LRU reference will prevent the buffer from being dropped. This means that the buffer can't actually be dropped until it is selected for eviction. There's no bound on the time spent on the LRU list, which means that the buffer may be undroppable for very long periods of time. Given that migration involves dropping buffers, the associated page is now unmigratible for long periods of time as well. CMA relies on being able to migrate a specific range of pages, so these these types of failures make CMA significantly less reliable, especially under high filesystem usage. Rather than waiting for the LRU algorithm to eventually kick out the buffer, explicitly remove the buffer from the LRU list when trying to drop it. There is still the possibility that the buffer could be added back on the list, but that indicates the buffer is still in use and would probably have other 'in use' indicates to prevent dropping. Signed-off-by: Laura Abbott <lauraa(a)codeaurora.org> --- fs/buffer.c | 38 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 38 insertions(+) diff --git a/fs/buffer.c b/fs/buffer.c index ad5938c..daa0c3d 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -1399,12 +1399,49 @@ static bool has_bh_in_lru(int cpu, void *dummy) return 0; } +static void __evict_bh_lru(void *arg) +{ + struct bh_lru *b = &get_cpu_var(bh_lrus); + struct buffer_head *bh = arg; + int i; + + for (i = 0; i < BH_LRU_SIZE; i++) { + if (b->bhs[i] == bh) { + brelse(b->bhs[i]); + b->bhs[i] = NULL; + goto out; + } + } +out: + put_cpu_var(bh_lrus); +} + +static bool bh_exists_in_lru(int cpu, void *arg) +{ + struct bh_lru *b = per_cpu_ptr(&bh_lrus, cpu); + struct buffer_head *bh = arg; + int i; + + for (i = 0; i < BH_LRU_SIZE; i++) { + if (b->bhs[i] == bh) + return 1; + } + + return 0; + +} void invalidate_bh_lrus(void) { on_each_cpu_cond(has_bh_in_lru, invalidate_bh_lru, NULL, 1, GFP_KERNEL); } EXPORT_SYMBOL_GPL(invalidate_bh_lrus); +void evict_bh_lrus(struct buffer_head *bh) +{ + on_each_cpu_cond(bh_exists_in_lru, __evict_bh_lru, bh, 1, GFP_ATOMIC); +} +EXPORT_SYMBOL_GPL(evict_bh_lrus); + void set_bh_page(struct buffer_head *bh, struct page *page, unsigned long offset) { @@ -3052,6 +3089,7 @@ drop_buffers(struct page *page, struct buffer_head **buffers_to_free) bh = head; do { + evict_bh_lrus(bh); if (buffer_write_io_error(bh) && page->mapping) set_bit(AS_EIO, &page->mapping->flags); if (buffer_busy(bh)) -- 1.7.11.3

12 years, 8 months

2
2
0 0

[GIT PULL] One more DMA-mapping fix for v3.6

by Marek Szyprowski

Hi Linus, I would like to ask for pulling one more patch for ARM dma-mapping subsystem to Linux v3.6 kernel tree. This patch fixes very subtle bug (typical off-by-one error) which might appear in very rare circumstances. The following changes since commit 55d512e245bc7699a8800e23df1a24195dd08217: Linux 3.6-rc5 (2012-09-08 16:43:45 -0700) are available in the git repository at: git://git.linaro.org/people/mszyprowski/linux-dma-mapping.git fixes-for-3.6 for you to fetch changes up to f3d87524975f01b885fc3d009c6ab6afd0d00746: arm: mm: fix DMA pool affiliation check (2012-09-10 16:15:48 +0200) Thanks! Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: ---------------------------------------------------------------- Thomas Petazzoni (1): arm: mm: fix DMA pool affiliation check arch/arm/mm/dma-mapping.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)

12 years, 8 months

1
0
0 0

Re: [Linaro-mm-sig] Converting OMAP's custom vram allocator

by Rob Clark

On Wed, Sep 5, 2012 at 5:08 AM, Tomi Valkeinen <tomi.valkeinen(a)ti.com> wrote: > Hi, > > OMAP has a custom video ram allocator, which I'd like to remove and use > the standard dma allocation functions. > > There are two problems for which I'd like to hear suggestions or > comments: > > First one is that the dma_alloc_* functions map the allocated memory for > cpu use. In many cases with OMAP DSS (display subsystem) this is not > needed: the memory may be written only by the SGX or the DSP, and it's > only read by the DSS, so it's never touched by the CPU. see dma_alloc_attrs() and DMA_ATTR_NO_KERNEL_MAPPING > This is even more true when using VRFB on omap3 (and probably TILER on > omap4) for rotation, as VRFB hides the actual memory and offers rotated > views. In this case the backend memory is never accessed by anyone else > than VRFB. just fwiw, we don't actually need contiguous memory on o4/tiler :-) (well, at least if you ignore things like secure playback) > Is there a way to allocate the memory without creating a mapping? While > it won't break anything as such, the allocated areas can be quite large > thus causing large areas of the kernel's memory space to be needlessly > reserved. > > The second case is passing a framebuffer address from the bootloader to > the kernel. Often with mobile devices the bootloader will initialize the > display hardware, showing a company logo or such. To keep the image on > the screen when kernel starts we need to reserve the same physical > memory area early at boot, and use that for the framebuffer. with a bit of handwaving, this is possible. You can pass a base address to dma_declare_contiguous() when you setup your device's CMA pool. Although that doesn't really guarantee you're allocation from that pool is at offset zero, I suppose. > I'm not sure if there's any actual problem with this one, presuming > there is a solution for the first case. Somehow the memory is reserved > at early boot time, and this is passed to the fb driver. But can the > memory be managed the same way as in normal case (for example freeing > it), or does it need to be handled as a special case? special-casing it might be better.. although possibly a dma attr could be added for this to tell dma_alloc_from_contiguous() that we need a particular address within the CMA pool. It seems a bit like a hack, but OTOH I guess pretty much every consumer device would need a hack like this. BR, -R > Tomi >

12 years, 9 months

1
0
0 0

[PATCH v3 0/2]gpu: ion: oom killer

by Zhangfei Gao

v2->v3 Split oom killer patch only. Based on Nishanth's patch, which change ion_debug_heap_total with id. 1. add heap_found 2. Solve the issue about serveral id share one type. Use ion_debug_heap_total(client, heap->id) instead of ion_debug_heap_total(client, heap->type) since id is unique while type can be shared. Fortunately Nishanth has update one patch, so rebase on the patch v1->v2 Sync to Aug 30 common.git v0->v1: 1. move ion_shrink out of mutex, suggested by Nishanth 2. check error flag of ERR_PTR(-ENOMEM) 3. add msleep to allow schedule out. Base on common.git, android-3.4 branch Add oom killer. Once heap is used off, SIGKILL is send to all tasks refered the buffer with descending oom_socre_adj Nishanth Peethambaran (1): gpu: ion: Update debugfs to show for each id Zhangfei Gao (1): gpu: ion: oom killer drivers/gpu/ion/ion.c | 131 +++++++++++++++++++++++++++++++++++++++++++++---- 1 files changed, 121 insertions(+), 10 deletions(-)

12 years, 9 months

1
2
0 0

[PATCH v2 0/3]gpu: ion: oom killer

by Zhangfei Gao

v1->v2 Sync to Aug 30 common.git v0->v1: 1, Change gen_pool_create(12, -1) to gen_pool_create(PAGE_SHIFT, -1), suggested by Haojian 2. move ion_shrink out of mutex, suggested by Nishanth 3. check error flag of ERR_PTR(-ENOMEM) 4. add msleep to allow schedule out. Base on common.git, android-3.4 branch Patch 2: Add support page wised cache flush for carveout_heap There is only one nents for carveout heap, as well as dirty bit. As a result, cache flush only takes effect for total carve heap. Patch 3: Add oom killer. Once heap is used off, SIGKILL is send to all tasks refered the buffer with descending oom_socre_adj Zhangfei Gao (3): gpu: ion: update carveout_heap_ops gpu: ion: carveout_heap page wised cache flush gpu: ion: oom killer drivers/gpu/ion/ion.c | 118 +++++++++++++++++++++++++++++++++- drivers/gpu/ion/ion_carveout_heap.c | 25 ++++++-- 2 files changed, 133 insertions(+), 10 deletions(-)

12 years, 9 months

3
12
0 0

[PATCH 1/2] gpu: ion: Fix kmalloc leak in carveout heap

by Nishanth Peethambaran

>From 78c843052f9438a848dd91334c6392b3b861bfa1 Mon Sep 17 00:00:00 2001 From: Nishanth Peethambaran <nishanth(a)broadcom.com> Date: Wed, 5 Sep 2012 02:21:54 +0530 Subject: [PATCH 1/2] gpu: ion: Fix kmalloc leak in carveout heap In carveout_unmap_dma(), the sg_table is freed. Change-Id: I8c2372c01a3f4c9061a1756ac9b31e09cf3ab9ef Signed-off-by: Nishanth Peethambaran <nishanth(a)broadcom.com> --- drivers/gpu/ion/ion_carveout_heap.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/drivers/gpu/ion/ion_carveout_heap.c b/drivers/gpu/ion/ion_carveout_heap.c index 5b6255b..f7608bc 100644 --- a/drivers/gpu/ion/ion_carveout_heap.c +++ b/drivers/gpu/ion/ion_carveout_heap.c @@ -107,6 +107,7 @@ void ion_carveout_heap_unmap_dma(struct ion_heap *heap, struct ion_buffer *buffer) { sg_free_table(buffer->sg_table); + kfree(buffer->sg_table); } void *ion_carveout_heap_map_kernel(struct ion_heap *heap, -- 1.7.0.4

12 years, 9 months

1
1
0 0

[PATCH 2/2] gpu: ion: Fix a memory leak

by Nishanth Peethambaran

>From 03184abb0efd05a422c4525f2f4398f127a9a0c9 Mon Sep 17 00:00:00 2001 From: Nishanth Peethambaran <nishanth(a)broadcom.com> Date: Wed, 5 Sep 2012 02:49:30 +0530 Subject: [PATCH 2/2] gpu: ion: Fix a memory leak The dirty bitmap buffer for cached buffers is freed when buffer gets destroyed. Change-Id: If2b6a6e3ffc5ed57623cfcbc01cb6b720bee532f Signed-off-by: Nishanth Peethambaran <nishanth(a)broadcom.com> --- drivers/gpu/ion/ion.c | 16 ++++++++++------ 1 files changed, 10 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/ion/ion.c b/drivers/gpu/ion/ion.c index 207d00f..3ad766a 100644 --- a/drivers/gpu/ion/ion.c +++ b/drivers/gpu/ion/ion.c @@ -176,12 +176,14 @@ static struct ion_buffer *ion_buffer_create(struct ion_heap *heap, return ERR_PTR(-EINVAL); } - ret = ion_buffer_alloc_dirty(buffer); - if (ret) { - heap->ops->unmap_dma(heap, buffer); - heap->ops->free(buffer); - kfree(buffer); - return ERR_PTR(ret); + if (buffer->flags & ION_FLAG_CACHED) { + ret = ion_buffer_alloc_dirty(buffer); + if (ret) { + heap->ops->unmap_dma(heap, buffer); + heap->ops->free(buffer); + kfree(buffer); + return ERR_PTR(ret); + } } buffer->dev = dev; @@ -210,6 +212,8 @@ static void ion_buffer_destroy(struct kref *kref) if (WARN_ON(buffer->kmap_cnt > 0)) buffer->heap->ops->unmap_kernel(buffer->heap, buffer); + if (buffer->flags & ION_FLAG_CACHED) + kfree(buffer->dirty); buffer->heap->ops->unmap_dma(buffer->heap, buffer); buffer->heap->ops->free(buffer); mutex_lock(&dev->lock); -- 1.7.0.4

12 years, 9 months

1
0
0 0

A few questions about the best way to implement RandR 1.4 / PRIME buffer sharing

by Aaron Plattner

So I've been experimenting with support for Dave Airlie's new RandR 1.4 provider object interface, so that Optimus-based laptops can use our driver to drive the discrete GPU and display on the integrated GPU. The good news is that I've got a proof of concept working. During a review of the current code, we came up with a few concerns: 1. The output source is responsible for allocating the shared memory Right now, the X server calls CreatePixmap on the output source screen and then expects the output sink screen to be able to display from whatever memory the source allocates. Right now, the source has no mechanism for asking the sink what its requirements are for the surface. I'm using our own internal pitch alignment requirements and that seems to be good enough for the Intel device to scan out, but that could be pure luck. Does it make sense to add a mechanism for drivers to negotiate this with each other, or is it sufficient to just define a lowest common denominator format and if your hardware can't deal with that format, you just don't get to share buffers? One of my coworkers brought to my attention the fact that Tegra requires a specific pitch alignment, and cannot accommodate larger pitches. If other SoC designs have similar restrictions, we might need to add a handshake mechanism. 2. There's no fallback mechanism if sharing can't be negotiated If RandR fails to share a pixmap with the output sink screen, the whole modeset fails. This means you'll end up not seeing anything on the screen and you'll probably think your computer locked up. Should there be some sort of software copy fallback to ensure that something at least shows up on the display? 3. How should the memory be allocated? In the prototype I threw together, I'm allocating the shared memory using shm_open and then exporting that as a dma-buf file descriptor using an ioctl I added to the kernel, and then importing that memory back into our driver through dma_buf_attach & dma_buf_map_attachment. Does it make sense for user-space programs to be able to export shmfs files like that? Should that interface go in DRM / GEM / PRIME instead? Something else? I'm pretty unfamiliar with this kernel code so any suggestions would be appreciated. -- Aaron P.S. for those unfamiliar with PRIME: Dave Airlie added new support to the X Resize and Rotate extension version 1.4 to support offloading display and rendering to different drivers. PRIME is the DRM implementation in the kernel, layered on top of DMA-BUF, that implements the actual sharing of buffers between drivers. http://cgit.freedesktop.org/xorg/proto/randrproto/tree/randrproto.txt?id=ra… http://airlied.livejournal.com/75555.html - update on hotplug server http://airlied.livejournal.com/76078.html - randr 1.5 demo videos ----------------------------------------------------------------------------------- This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. -----------------------------------------------------------------------------------

12 years, 9 months

3
3
0 0

[PATCH 0/3] gpu: ion: oom killer

by Zhangfei Gao

Base on common.git, android-3.4 branch Patch 2: Add support page wised cache flush for carveout_heap There is only one nents for carveout heap, as well as dirty bit. As a result, cache flush only takes effect for total carve heap. Patch 3: Add oom killer. Once heap is used off, SIGKILL is send to all tasks refered the buffer with descending oom_socre_adj Zhangfei Gao (3): gpu: ion: update carveout_heap_ops gpu: ion: carveout_heap page wised cache flush gpu: ion: oom killer drivers/gpu/ion/ion.c | 112 ++++++++++++++++++++++++++++++++++- drivers/gpu/ion/ion_carveout_heap.c | 23 ++++++-- 2 files changed, 127 insertions(+), 8 deletions(-)

12 years, 9 months

4
17
0 0

[PATCH 0/3] gpu: ion: oom killer

by Zhangfei Gao

v0->v1: 1, Change gen_pool_create(12, -1) to gen_pool_create(PAGE_SHIFT, -1), suggested by Haojian 2. move ion_shrink out of mutex, suggested by Nishanth 3. check error flag of ERR_PTR(-ENOMEM) 4. add msleep to allow schedule out. Base on common.git, android-3.4 branch Patch 2: Add support page wised cache flush for carveout_heap There is only one nents for carveout heap, as well as dirty bit. As a result, cache flush only takes effect for total carve heap. Patch 3: Add oom killer. Once heap is used off, SIGKILL is send to all tasks refered the buffer with descending oom_socre_adj Zhangfei Gao (3): gpu: ion: update carveout_heap_ops gpu: ion: carveout_heap page wised cache flush gpu: ion: oom killer drivers/gpu/ion/ion.c | 118 +++++++++++++++++++++++++++++++++- drivers/gpu/ion/ion_carveout_heap.c | 25 ++++++-- 2 files changed, 133 insertions(+), 10 deletions(-)

12 years, 9 months

3
9
0 0

[GIT PULL] DMA-mapping fixes for v3.6

by Marek Szyprowski

Hi Linus, I would like to ask for pulling another set of fixes for ARM dma-mapping subsystem. Commit e9da6e9905e6 replaced custom consistent buffer remapping code with generic vmalloc areas. It however introduced some regressions caused by limited support for allocations in atomic context. This series contains fixes for those regressions. For some subplatforms the default, pre-allocated pool for atomic allocations turned out to be too small, so a function for setting its size has been added. Another set of patches adds support for atomic allocations to IOMMU-aware DMA-mapping implementation. The last part of this pull request contains two fixes for Contiguous Memory Allocator, which relax too strict requirements. The following changes since commit fea7a08acb13524b47711625eebea40a0ede69a0: Linux 3.6-rc3 (2012-08-22 13:29:06 -0700) are available in the git repository at: fixes-for-3.6 for you to fetch changes up to 479ed93a4b98eef03fd8260f7ddc00019221c450: ARM: dma-mapping: IOMMU allocates pages from atomic_pool with GFP_ATOMIC (2012-08-28 21:01:07 +0200) Thanks! Best regards Marek Szyprowski Samsung Poland R&D Center ---------------------------------------------------------------- Patch summary: Hiroshi Doyu (4): ARM: dma-mapping: atomic_pool with struct page **pages ARM: dma-mapping: Refactor out to introduce __in_atomic_pool ARM: dma-mapping: Introduce __atomic_get_pages() for __iommu_get_pages() ARM: dma-mapping: IOMMU allocates pages from atomic_pool with GFP_ATOMIC Marek Szyprowski (5): mm: cma: fix alignment requirements for contiguous regions ARM: relax conditions required for enabling Contiguous Memory Allocator ARM: DMA-Mapping: add function for setting coherent pool size from platform code ARM: DMA-Mapping: print warning when atomic coherent allocation fails ARM: Kirkwood: increase atomic coherent pool size arch/arm/Kconfig | 2 +- arch/arm/include/asm/dma-mapping.h | 7 ++ arch/arm/mach-kirkwood/common.c | 7 ++ arch/arm/mm/dma-mapping.c | 114 ++++++++++++++++++++++++++++++++--- drivers/base/dma-contiguous.c | 2 +- 5 files changed, 120 insertions(+), 12 deletions(-)

12 years, 9 months

1
0
0 0

Exporting ion buffer to multiple process from same driver

by Nishanth Peethambaran

Hi, I am trying to export an ion_buffer allocate from kernel space to multiple user-space clients. Eg: Allow multiple process to mmap framebuffer allocated using ion by fb driver. The following is the pseudo-code for that. Is this fine? there a cleaner way to do it? Or is it expected to share buffers across process only by user-space sharing fds using sockets/binder and not directly in kernel. fb driver init/probe: (init process context) ------------------------------------------------- /* Create an ion client and allocate framebuffer */ init_client = ion_client_create(idev,...); init_hdl = ion_alloc(init_client,...); /* Create a global dma_buf instance for the buffer */ fd = ion_share_dma_buf(init_client, init_hdl); // - Inc refcount of ion_buffer // - Create a dma_buf and anon-file for the ion buffer // - Get a free fd and install to anon file g_dma_buf = dma_buf_get(fd); // - Get the dma_buf pointer and inc refcount of anon_file dma_buf_put(g_dma_buf); // - Dec extra refcount of anon_file which happened in prev command put_unused_fd(fd); // - Free up the fd as fd is not exported to user-space here. fb driver exit: (init process context) ------------------------------------------ /* Free the dma_buf reference */ dma_buf_put(g_dma_buf); // - Dec refcount of anon_file. Free the dma_buf and dec refcount of ion_buffer if anon_file refcount = 0 /* Free the framebuffer and destroy the ion client created for init process */ ion_free(init_client, init_hdl); ion_client_destroy(init_client); fb device open: (user process context) ----------------------------------------------- /* Create an ion client for the user process */ p_client = ion_client_create(idev,...); fb device ioctl to import ion handle for the fb: (user process context) ----------------------------------------------------------------------------------- /* Import a ion_handle from the global dma_buf */ fd = dma_buf_fd(g_dmabuf, O_CLOEXEC); // - Get ref to anon file // - Get a free fd and install to anon file p_hdl = ion_import_dma_buf(p_client, fd); // - Inc refcount of ion_buffer // - create a ion_handle for the buffer for this process/client dma_buf_put(g_dmabuf); // - Free the anon file reference taken in first step put_unused_fd(fd); // - Free up the fd as fd is not exported to user-space here. fb device release: (user process context) --------------------------------------------------- /* Destroy the client created */ ion_client_destroy(p_client); - Nishanth Peethambaran

12 years, 9 months

1
0
0 0

CMA page migration failure due to buffers on bh_lru

by Laura Abbott

Hi, I've been observing a high rate of failures with CMA allocations on my ARM system. I've set up a test case set up with a 56MB CMA region that essentially does the following: total_failures = 0; loop forever: loop_failure = 0; for (i = 0; i < 56; i++) chunk[i] = dma_allocate(&cma_dev, 1MB) if (!chunk[i]) loop_failure = 0 if (loop_failure) total_failures++ loop_failure = 0 for (i = 0; i < 56; i++) dma_free(&cma_dev, chunk[i], 1MB) In the background, I also have a process doing some amount of filesystem activity (adb push/pull since this is an android system). During the course of my investigations I generally get ~8500 loops total and ~450 total failures (i.e. one or more buffers could not be allocated). This is unacceptably high for our use cases. In every case the allocation failure was ultimately due to a migration failure; the pages contained buffers which could not be dropped because the buffers were busy (move_to_new_page -> fallback_migrate_page -> try_to_release_page -> try_to_free_buffers -> drop_buffers -> buffer_busy). In every case, the b_count on the buffer head was always 1. The problem arises because of the LRU lists for buffer heads: __getblk __getblk_slow grow_buffers grow_dev_page find_or_create_page -- create a possibly movable page __find_get_block __find_get_block_slow find_get_page -- return the movable page bh_lru_install get_bh -- buffer head now has a reference The reference taken in bh_lru_install won't be dropped until the bh is evicted from the lru. This means the page cannot be migrated as long as the buffer exists on an LRU list. The real issue is that unless the buffer gets evicted quickly the page can remain non-migratible for long periods of time. This makes CMA regions unusable for long periods of time given that we generally don't want to size CMA regions any larger than necessary ergo any failure will cause a problem. My quick and dirty workaround for testing is to remove the GFP_MOVABLE flag from find_or_create_page but this seems significantly less than optimal. Ideally, it seems like the buffers should be evicted from the LRU when trying to drop (expand on invalid_bh_lru?) but I'm not familiar enough with the code path to know if this is a good approach. Any suggestions/feedback is appreciated. Thanks. Laura -- Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

12 years, 9 months

1
1
0 0

[linaro-mm-sig] Backward compatibility of 3.4 android kernel to ICS

by Nishanth Peethambaran

Hi, I see that the lowmemkiller.c is changed to use oom_score_adj instead of oom_adj. Does this mean I cannot use an ICS system image with 3.4 kernel? Or is there a workaround? - Nishanth Peethambaran

12 years, 9 months

1
0
0 0

Re: [Linaro-mm-sig] CMA allocation issue : some pages can't be migrated

by Aubertin, Guillaume

looping-in the linaro-mm-sig ML. On Thu, Aug 30, 2012 at 4:47 PM, Aubertin, Guillaume <g-aubertin(a)ti.com>wrote: > hi guys, > > I've been working for a few days on getting a proper rmmod with the > remoteproc/rpmsg modules, and I stumbled upon an interesting issue. > > when doing sucessive memory allocation and release in the CMA > reservation (by loading/unloading the firmware several times), the > following message shows up : > > [ 119.908477] cma: dma_alloc_from_contiguous(cma ed10ad00, count 256, > align 8) > [ 119.908843] cma: dma_alloc_from_contiguous(): memory range at c0dfb000 > is busy, retrying > [ 119.909698] cma: dma_alloc_from_contiguous(): returned c0dfd000 > > dma_alloc_from_contiguous() tries to allocate the following range, > 0xc0dfd000, succesfully this time. > > In some cases, the allocation fails after trying several ranges : > > [ 119.912231] cma: dma_alloc_from_contiguous(cma ed10ad00, count 768, > align 8) > [ 119.912719] cma: dma_alloc_from_contiguous(): memory range at c0dff000 > is busy, retrying > [ 119.913055] cma: dma_alloc_from_contiguous(): memory range at c0e01000 > is busy, retrying > [ 119.913055] rproc remoteproc0: dma_alloc_coherent failed: 3145728 > > Here is my understanding so far : > > First, even if we made a CMA reservation, the kernel can still allocate > pages in this area, but these pages must be movable (user process page by > example). > > When dma_alloc_from_contiguous() is called to allocate X pages, it looks > for the next X contiguous free pages in it's CMA bitmap (with respect to > the memory alignment). Then, alloc_contig_range() is called to allocate the > given range of pages. Alloc_contig_range() analyses the pages we want to > allocate, and if a page is already used, it is migrated to a new page > outside the page array we want to reserve. this is done using > isolate_migratepages_range() to list the pages to migrate, and > migrate_pages() to try to migrate the pages, and that's where it fails. > Below is a list of next function calls : > > fallback_migrate_page() --> migrate_page() --> try_to_release_page() > --> try_to_free_buffer() --> drop_buffers() --> buffer_busy() > > I understand here that the page contains used buffers that can't be > dropped, and so the page can't be migrated. Well, I must admit that once > here, I'm feeling a little lost in this ocean of memory management code ;). > After a few researches, I found the following thread on the > linux-arm-kernel ML talking about the same issue : > > http://lists.infradead.org/pipermail/linux-arm-kernel/2012-June/102844.html with > the following patch : > > * mm/page_alloc.c | 3 ++-* > * 1 files changed, 2 insertions(+), 1 deletions(-)* > * > * > *diff --git a/mm/page_alloc.c b/mm/page_alloc.c* > *index 0e1c6f5..c9a6483 100644* > *--- a/mm/page_alloc.c* > *+++ b/mm/page_alloc.c* > *@@ -1310,7 +1310,8 @@ void free_hot_cold_page(struct page *page, int > cold)* > * * excessively into the page allocator* > * */* > * if (migratetype >= MIGRATE_PCPTYPES) {* > *- if (unlikely(migratetype == MIGRATE_ISOLATE)) {* > *+ if (unlikely(migratetype == MIGRATE_ISOLATE)* > *+ || is_migrate_cma(migratetype)) {* > * free_one_page(zone, page, 0, migratetype);* > * goto out;* > * }* > > I tried the patch, and it seems to work (I didn't have any "memory range > busy" in 5000+ tests), but I'm affraid that this could have some nasty side > effects. > > Any idea ? > > Thanks in advance, > Guillaume > > > -- > Texas Instruments France SA, 821 Avenue Jack Kilby, 06270 Villeneuve > Loubet. 036 420 040 R.C.S Antibes. Capital de EUR 753.920 > -- Texas Instruments France SA, 821 Avenue Jack Kilby, 06270 Villeneuve Loubet. 036 420 040 R.C.S Antibes. Capital de EUR 753.920

12 years, 9 months

2
2
0 0

[RFC] New dma_buf -> EGLImage EGL extension

by Tom Cooksey

Hi All, Over the last few months I've been working on & off with a few people from Linaro on a new EGL extension. The extension allows constructing an EGLImage from a (set of) dma_buf file descriptors, including support for multi-plane YUV. I envisage the primary use-case of this extension to be importing video frames from v4l2 into the EGL/GLES graphics driver to texture from. Originally the intent was to develop this as a Khronos-ratified extension. However, this is a little too platform-specific to be an officially sanctioned Khronos extension. It also goes against the general "EGLStream" direction the EGL working group is going in. As such, the general feeling was to make this an EXT "multi-vendor" extension with no official stamp of approval from Khronos. As this is no-longer intended to be a Khronos extension, I've re-written it to be a lot more Linux & dma_buf specific. It also allows me to circulate the extension more widely (I.e. To those outside Khronos membership). ARM are implementing this extension for at least our Mali-T6xx driver and likely earlier drivers too. I am sending this e-mail to solicit feedback, both from other vendors who might implement this extension (Mesa3D?) and from potential users of the extension. However, any feedback is welcome. Please find the extension text as it currently stands below. There several open issues which I've proposed solutions for, but I'm not really happy with those proposals and hoped others could chip-in with better ideas. There are likely other issues I've not thought about which also need to be added and addressed. Once there's a general consensus or if no-one's interested, I'll update the spec, move it out of Draft status and get it added to the Khronos registry, which includes assigning values for the new symbols. Cheers, Tom ---------8<--------- Name EXT_image_dma_buf_import Name Strings EGL_EXT_image_dma_buf_import Contributors Jesse Barker Rob Clark Tom Cooksey Contacts Jesse Barker (jesse 'dot' barker 'at' linaro 'dot' org) Tom Cooksey (tom 'dot' cooksey 'at' arm 'dot' com) Status DRAFT Version Version 3, August 16, 2012 Number EGL Extension ??? Dependencies EGL 1.2 is required. EGL_KHR_image_base is required. The EGL implementation must be running on a Linux kernel supporting the dma_buf buffer sharing mechanism. This extension is written against the wording of the EGL 1.2 Specification. Overview This extension allows creating an EGLImage from a Linux dma_buf file descriptor or multiple file descriptors in the case of multi-plane YUV images. New Types None New Procedures and Functions None New Tokens Accepted by the <target> parameter of eglCreateImageKHR: EGL_LINUX_DMA_BUF_EXT Accepted as an attribute in the <attrib_list> parameter of eglCreateImageKHR: EGL_LINUX_DRM_FOURCC_EXT EGL_DMA_BUF_PLANE0_FD_EXT EGL_DMA_BUF_PLANE0_OFFSET_EXT EGL_DMA_BUF_PLANE0_PITCH_EXT EGL_DMA_BUF_PLANE1_FD_EXT EGL_DMA_BUF_PLANE1_OFFSET_EXT EGL_DMA_BUF_PLANE1_PITCH_EXT EGL_DMA_BUF_PLANE2_FD_EXT EGL_DMA_BUF_PLANE2_OFFSET_EXT EGL_DMA_BUF_PLANE2_PITCH_EXT Additions to Chapter 2 of the EGL 1.2 Specification (EGL Operation) Add to section 2.5.1 "EGLImage Specification" (as defined by the EGL_KHR_image_base specification), in the description of eglCreateImageKHR: "Values accepted for <target> are listed in Table aaa, below. +-------------------------+--------------------------------------------+ | <target> | Notes | +-------------------------+--------------------------------------------+ | EGL_LINUX_DMA_BUF_EXT | Used for EGLImages imported from Linux | | | dma_buf file descriptors | +-------------------------+--------------------------------------------+ Table aaa. Legal values for eglCreateImageKHR <target> parameter ... If <target> is EGL_LINUX_DMA_BUF_EXT, <dpy> must be a valid display, <ctx> must be EGL_NO_CONTEXT, and <buffer> must be NULL, cast into the type EGLClientBuffer. The details of the image is specified by the attributes passed into eglCreateImageKHR. Required attributes and their values are as follows: * EGL_WIDTH & EGL_HEIGHT: The logical dimensions of the buffer in pixels * EGL_LINUX_DRM_FOURCC_EXT: The pixel format of the buffer, as specified by drm_fourcc.h and used as the pixel_format parameter of the drm_mode_fb_cmd2 ioctl. * EGL_DMA_BUF_PLANE0_FD_EXT: The dma_buf file descriptor of plane 0 of the image. * EGL_DMA_BUF_PLANE0_OFFSET_EXT: The offset from the start of the dma_buf of the first sample in plane 0, in bytes. * EGL_DMA_BUF_PLANE0_PITCH_EXT: The number of bytes between the start of subsequent rows of samples in plane 0. May have special meaning for non-linear formats. For images in an RGB color-space or those using a single-plane YUV format, only the first plane's file descriptor, offset & pitch should be specified. For semi-planar YUV formats, the chroma samples are stored in plane 1 and for fully planar formats, U-samples are stored in plane 1 and V-samples are stored in plane 2. Planes 1 & 2 are specified by the following attributes, which have the same meanings as defined above for plane 0: * EGL_DMA_BUF_PLANE1_FD_EXT * EGL_DMA_BUF_PLANE1_OFFSET_EXT * EGL_DMA_BUF_PLANE1_PITCH_EXT * EGL_DMA_BUF_PLANE2_FD_EXT * EGL_DMA_BUF_PLANE2_OFFSET_EXT * EGL_DMA_BUF_PLANE2_PITCH_EXT If eglCreateImageKHR is successful for a EGL_LINUX_DMA_BUF_EXT target, the EGL takes ownership of the file descriptor and is responsible for closing it, which it may do at any time while the EGLDisplay is initialized." Add to the list of error conditions for eglCreateImageKHR: "* If <target> is EGL_LINUX_DMA_BUF_EXT and <buffer> is not NULL, the error EGL_BAD_PARAMETER is generated. * If <target> is EGL_LINUX_DMA_BUF_EXT, and the list of attributes is incomplete, EGL_BAD_PARAMETER is generated. * If <target> is EGL_LINUX_DMA_BUF_EXT, and the EGL_LINUX_DRM_FOURCC_EXT attribute is set to a format not supported by the EGL, EGL_BAD_MATCH is generated. * If <target> is EGL_LINUX_DMA_BUF_EXT, and the EGL_LINUX_DRM_FOURCC_EXT attribute indicates a single-plane format, EGL_BAD_ATTRIBUTE is generated if any of the EGL_DMA_BUF_PLANE1_* or EGL_DMA_BUF_PLANE2_* attributes are specified. Issues 1. Should this be a KHR or EXT extension? ANSWER: EXT. Khronos EGL working group not keen on this extension as it is seen as contradicting the EGLStream direction the specification is going in. The working group recommends creating additional specs to allow an EGLStream producer/consumer connected to v4l2/DRM or any other Linux interface. 2. Should this be a generic any platform extension, or a Linux-only extension which explicitly states the handles are dma_buf fds? ANSWER: There's currently no intention to port this extension to any OS not based on the Linux kernel. Consequently, this spec can be explicitly written against Linux and the dma_buf API. 3. Does ownership of the file descriptor pass to the EGL library? PROPOSAL: If eglCreateImageKHR is successful, EGL assumes ownership of the file descriptors and is responsible for closing them. 4. How are the different YUV color spaces handled (BT.709/BT.601)? Open issue, still TBD. Doesn't seem to be specified by either the v4l2 or DRM APIs. PROPOSAL: Undefined and implementation/format dependent. 5. What chroma-siting is used for sub-sampled YUV formats? Open issue, still TBD. Doesn't seem to be specified by either the v4l2 or DRM APIs. PROPOSAL: Undefined and implementation/format dependent. 5. How can an application query which formats the EGL implementation supports? PROPOSAL: Don't provide a query mechanism but instead add an error condition that EGL_BAD_MATCH is raised if the EGL implementation doesn't support that particular format. 5. Which image formats should be supported and how is format specified? Open issue, still TBD. Seem to be two options 1) specify a new enum in this specification and enumerate all possible formats. 2) Use an existing enum already in Linux, either v4l2_mbus_pixelcode and/or those formats listed in drm_fourcc.h? PROPOSAL: Go for option 2) and just use values defined in drm_fourcc.h. Revision History #3 (Tom Cooksey, August 16, 2012) - Changed name from EGL_EXT_image_external and re-written language to explicitly state this for use with Linux & dma_buf. - Added a list of issues, including some still open ones. #2 (Jesse Barker, May 30, 2012) - Revision to split eglCreateImageKHR functionality from export Functionality. - Update definition of EGLNativeBufferType to be a struct containing a list of handles to support multi-buffer/multi-planar formats. #1 (Jesse Barker, March 20, 2012) - Initial draft.

12 years, 9 months

1
0
0 0

[PATCH 1/4] dma-buf: remove fallback for !CONFIG_DMA_SHARED_BUFFER

by Maarten Lankhorst

Documentation says that code requiring dma-buf should add it to select, so inline fallbacks are not going to be used. A link error will make it obvious what went wrong, instead of silently doing nothing at runtime. Signed-off-by: Maarten Lankhorst <maarten.lankhorst(a)canonical.com> --- include/linux/dma-buf.h | 99 ----------------------------------------------- 1 file changed, 99 deletions(-) diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h index eb48f38..bd2e52c 100644 --- a/include/linux/dma-buf.h +++ b/include/linux/dma-buf.h @@ -156,7 +156,6 @@ static inline void get_dma_buf(struct dma_buf *dmabuf) get_file(dmabuf->file); } -#ifdef CONFIG_DMA_SHARED_BUFFER struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf, struct device *dev); void dma_buf_detach(struct dma_buf *dmabuf, @@ -184,103 +183,5 @@ int dma_buf_mmap(struct dma_buf *, struct vm_area_struct *, unsigned long); void *dma_buf_vmap(struct dma_buf *); void dma_buf_vunmap(struct dma_buf *, void *vaddr); -#else - -static inline struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf, - struct device *dev) -{ - return ERR_PTR(-ENODEV); -} - -static inline void dma_buf_detach(struct dma_buf *dmabuf, - struct dma_buf_attachment *dmabuf_attach) -{ - return; -} - -static inline struct dma_buf *dma_buf_export(void *priv, - const struct dma_buf_ops *ops, - size_t size, int flags) -{ - return ERR_PTR(-ENODEV); -} - -static inline int dma_buf_fd(struct dma_buf *dmabuf, int flags) -{ - return -ENODEV; -} - -static inline struct dma_buf *dma_buf_get(int fd) -{ - return ERR_PTR(-ENODEV); -} - -static inline void dma_buf_put(struct dma_buf *dmabuf) -{ - return; -} - -static inline struct sg_table *dma_buf_map_attachment( - struct dma_buf_attachment *attach, enum dma_data_direction write) -{ - return ERR_PTR(-ENODEV); -} - -static inline void dma_buf_unmap_attachment(struct dma_buf_attachment *attach, - struct sg_table *sg, enum dma_data_direction dir) -{ - return; -} - -static inline int dma_buf_begin_cpu_access(struct dma_buf *dmabuf, - size_t start, size_t len, - enum dma_data_direction dir) -{ - return -ENODEV; -} - -static inline void dma_buf_end_cpu_access(struct dma_buf *dmabuf, - size_t start, size_t len, - enum dma_data_direction dir) -{ -} - -static inline void *dma_buf_kmap_atomic(struct dma_buf *dmabuf, - unsigned long pnum) -{ - return NULL; -} - -static inline void dma_buf_kunmap_atomic(struct dma_buf *dmabuf, - unsigned long pnum, void *vaddr) -{ -} - -static inline void *dma_buf_kmap(struct dma_buf *dmabuf, unsigned long pnum) -{ - return NULL; -} - -static inline void dma_buf_kunmap(struct dma_buf *dmabuf, - unsigned long pnum, void *vaddr) -{ -} - -static inline int dma_buf_mmap(struct dma_buf *dmabuf, - struct vm_area_struct *vma, - unsigned long pgoff) -{ - return -ENODEV; -} - -static inline void *dma_buf_vmap(struct dma_buf *dmabuf) -{ - return NULL; -} - -static inline void dma_buf_vunmap(struct dma_buf *dmabuf, void *vaddr) -{ -} -#endif /* CONFIG_DMA_SHARED_BUFFER */ #endif /* __DMA_BUF_H__ */

12 years, 9 months

5
22
0 0

[PATCH 1/4] ARM: dma-mapping: Small logical clean up

by Hiroshi Doyu

Skip unnecessary operations if order == 0. A little bit easier to read. Signed-off-by: Hiroshi Doyu <hdoyu(a)nvidia.com> --- arch/arm/mm/dma-mapping.c | 9 +++++---- 1 files changed, 5 insertions(+), 4 deletions(-) diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c index 70a6275..aaea5e4 100644 --- a/arch/arm/mm/dma-mapping.c +++ b/arch/arm/mm/dma-mapping.c @@ -1032,11 +1032,12 @@ static struct page **__iommu_alloc_buffer(struct device *dev, size_t size, gfp_t if (!pages[i]) goto error; - if (order) + if (order) { split_page(pages[i], order); - j = 1 << order; - while (--j) - pages[i + j] = pages[i] + j; + j = 1 << order; + while (--j) + pages[i + j] = pages[i] + j; + } __dma_clear_buffer(pages[i], PAGE_SIZE << order); i += 1 << order; -- 1.7.5.4

12 years, 9 months

2
5
0 0

[PATCH 0/3] Add code for setting coherent pool size

by Marek Szyprowski

Hi! Aaro Koskinen and Josh Coombs reported that commit e9da6e9905e639 ("ARM: dma-mapping: remove custom consistent dma region") introduced a regresion. It turned out that the default 256KiB for atomic coherent pool might not be enough. After that patch, some Kirkwood systems run out of atomic coherent memory and fail without any meanfull message. This patch series is an attempt to fix those issues by adding function for setting coherent pool size from platform initialization code and increasing the size of the pool for Kirkwood systems. Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: Marek Szyprowski (3): ARM: DMA-Mapping: add function for setting coherent pool size from platform code ARM: DMA-Mapping: print warning when atomic coherent allocation fails ARM: Kirkwood: increase atomic coherent pool size arch/arm/include/asm/dma-mapping.h | 7 +++++++ arch/arm/mach-kirkwood/common.c | 7 +++++++ arch/arm/mm/dma-mapping.c | 22 +++++++++++++++++++++- 3 files changed, 35 insertions(+), 1 deletions(-) -- 1.7.1.569.g6f426

12 years, 9 months

4
12
0 0

[v4 0/4] ARM: dma-mapping: IOMMU atomic allocation

by Hiroshi Doyu

Hi, The commit e9da6e9 "ARM: dma-mapping: remove custom consistent dma region" breaks the compatibility with existing drivers. This causes the following kernel oops(*1). That driver has called dma_pool_alloc() to allocate memory from the interrupt context, and it hits BUG_ON(in_interrpt()) in "get_vm_area_caller()". This patch seris fixes this problem with making use of the pre-allocate atomic memory pool which DMA is using in the same way as DMA does now. Any comment would be really appreciated. v4: Fix plain memory allocation. (Konrad,Marek) Print nicer error message at __in_atomic_pool() (Konrad) v3: Provide a different path for IOMMU for more clean code. (Marek) atomic_pool is backed with struct page *pages[]. http://lists.linaro.org/pipermail/linaro-mm-sig/2012-August/002446.html v2: Don't modify attrs(DMA_ATTR_NO_KERNEL_MAPPING) for atomic allocation. (Marek) Modify vzalloc (KyongHo,Minchan) http://lists.linaro.org/pipermail/linaro-mm-sig/2012-August/002430.html v1: http://lists.linaro.org/pipermail/linaro-mm-sig/2012-August/002398.html *1: [ 8.321343] ------------[ cut here ]------------ [ 8.325971] kernel BUG at kernel/mm/vmalloc.c:1322! [ 8.333615] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ARM [ 8.339436] Modules linked in: [ 8.342496] CPU: 0 Tainted: G W (3.4.6-00067-g5d485f7 #67) [ 8.349192] PC is at __get_vm_area_node.isra.29+0x164/0x16c [ 8.354758] LR is at get_vm_area_caller+0x4c/0x54 [ 8.359454] pc : [<c011297c>] lr : [<c011318c>] psr: 20000193 [ 8.359458] sp : c09edca0 ip : c09ec000 fp : ae278000 [ 8.370922] r10: f0000000 r9 : c011aa54 r8 : c0a26cb8 [ 8.376136] r7 : 00000001 r6 : 000000d0 r5 : 20000008 r4 : c09edca0 [ 8.382651] r3 : 00010000 r2 : 20000008 r1 : 00000001 r0 : 00001000 [ 8.389166] Flags: nzCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel [ 8.396549] Control: 10c5387d Table: ad98c04a DAC: 00000015 .... [ 9.169162] dfa0: 412fc099 c09ec000 00000000 c000fdd8 c06df1e4 c0a1b080 00000000 00000000 [ 9.177329] dfc0: c0a235cc 8000406a 00000000 c0986818 ffffffff ffffffff c0986404 00000000 [ 9.185497] dfe0: 00000000 c09bb070 10c5387d c0a19c58 c09bb064 80008044 00000000 00000000 [ 9.193673] [<c011297c>] (__get_vm_area_node.isra.29+0x164/0x16c) from [<c011318c>] (get_vm_area_caller+0x4c/0x54) [ 9.204022] [<c011318c>] (get_vm_area_caller+0x4c/0x54) from [<c001aed8>] (__iommu_alloc_remap.isra.14+0x2c/0xfc) [ 9.214276] [<c001aed8>] (__iommu_alloc_remap.isra.14+0x2c/0xfc) from [<c001b06c>] (arm_iommu_alloc_attrs+0xc4/0xf8) [ 9.224795] [<c001b06c>] (arm_iommu_alloc_attrs+0xc4/0xf8) from [<c011aa54>] (pool_alloc_page.constprop.5+0x6c/0xf8) [ 9.235309] [<c011aa54>] (pool_alloc_page.constprop.5+0x6c/0xf8) from [<c011ab60>] (dma_pool_alloc+0x80/0x170) [ 9.245304] [<c011ab60>] (dma_pool_alloc+0x80/0x170) from [<c03cbbcc>] (tegra_build_dtd+0x48/0x14c) [ 9.254344] [<c03cbbcc>] (tegra_build_dtd+0x48/0x14c) from [<c03cbd4c>] (tegra_req_to_dtd+0x7c/0xa8) [ 9.263467] [<c03cbd4c>] (tegra_req_to_dtd+0x7c/0xa8) from [<c03cc140>] (tegra_ep_queue+0x154/0x33c) [ 9.272592] [<c03cc140>] (tegra_ep_queue+0x154/0x33c) from [<c03dd5b4>] (composite_setup+0x364/0x6d4) [ 9.281804] [<c03dd5b4>] (composite_setup+0x364/0x6d4) from [<c03dd9dc>] (android_setup+0xb8/0x14c) [ 9.290843] [<c03dd9dc>] (android_setup+0xb8/0x14c) from [<c03cd144>] (setup_received_irq+0xbc/0x270) [ 9.300053] [<c03cd144>] (setup_received_irq+0xbc/0x270) from [<c03cda64>] (tegra_udc_irq+0x2ac/0x2c4) [ 9.309353] [<c03cda64>] (tegra_udc_irq+0x2ac/0x2c4) from [<c00b5708>] (handle_irq_event_percpu+0x78/0x2e0) [ 9.319087] [<c00b5708>] (handle_irq_event_percpu+0x78/0x2e0) from [<c00b59b4>] (handle_irq_event+0x44/0x64) [ 9.328907] [<c00b59b4>] (handle_irq_event+0x44/0x64) from [<c00b8688>] (handle_fasteoi_irq+0xc4/0x16c) [ 9.338294] [<c00b8688>] (handle_fasteoi_irq+0xc4/0x16c) from [<c00b4f14>] (generic_handle_irq+0x34/0x48) [ 9.347858] [<c00b4f14>] (generic_handle_irq+0x34/0x48) from [<c000f6f4>] (handle_IRQ+0x54/0xb4) [ 9.356637] [<c000f6f4>] (handle_IRQ+0x54/0xb4) from [<c00084b0>] (gic_handle_irq+0x2c/0x60) [ 9.365068] [<c00084b0>] (gic_handle_irq+0x2c/0x60) from [<c000e900>] (__irq_svc+0x40/0x70) [ 9.373405] Exception stack(0xc09edf10 to 0xc09edf58) [ 9.378447] df00: 00000000 000f4240 00000003 00000000 [ 9.386615] df20: 00000000 e55bbc00 ef66f3ca 00000001 00000000 412fc099 c0abb9c8 00000000 [ 9.394781] df40: 3b9ac9ff c09edf58 c027a9bc c0042880 20000113 ffffffff [ 9.401396] [<c000e900>] (__irq_svc+0x40/0x70) from [<c0042880>] (tegra_idle_enter_lp3+0x68/0x78) [ 9.410272] [<c0042880>] (tegra_idle_enter_lp3+0x68/0x78) from [<c04701d4>] (cpuidle_idle_call+0xdc/0x3a4) [ 9.419922] [<c04701d4>] (cpuidle_idle_call+0xdc/0x3a4) from [<c000fdd8>] (cpu_idle+0xd8/0x134) [ 9.428612] [<c000fdd8>] (cpu_idle+0xd8/0x134) from [<c0986818>] (start_kernel+0x27c/0x2cc) [ 9.436952] Code: e1a00004 e3a04000 eb002265 eaffffe0 (e7f001f2) [ 9.443038] ---[ end trace 1b75b31a2719ed24 ]--- [ 9.447645] Kernel panic - not syncing: Fatal exception in interrupt Hiroshi Doyu (4): ARM: dma-mapping: atomic_pool with struct page **pages ARM: dma-mapping: Refactor out to introduce __in_atomic_pool ARM: dma-mapping: Introduce __atomic_get_pages() for __iommu_get_pages() ARM: dma-mapping: IOMMU allocates pages from atomic_pool with GFP_ATOMIC arch/arm/mm/dma-mapping.c | 91 ++++++++++++++++++++++++++++++++++++++++---- 1 files changed, 82 insertions(+), 9 deletions(-) -- 1.7.5.4

12 years, 9 months

2
5
0 0

Question about ION carveout heap support partial cache flush

by zhangfei gao

Hi, All We met question about dmac_map_area & dmac_flush_range from user addr. mcr would not return on armv7 processor. Existing ion carveout heap does not support partial cache flush. Total cache will be flushed at all. There is only one dirty bit for carveout heap, as well as sg_table->nents. drivers/gpu/ion/ion_carveout_heap.c ion_carveout_heap_map_dma -> sg_alloc_table(table, 1, GFP_KERNEL); ion_buffer_alloc_dirty -> pages = buffer->sg_table->nents; We want to support partial cache flush. Align to cache line, instead of PAGE_SIZE, for efficiency consideration. We have considered extended dirty bit, but looks like only align to PAGE_SIZE. For experiment we modify ioctl ION_IOC_SYNC on armv7. And directly use dmac_map_area & dmac_flush_range with add from user space. However, we find dmac_map_area can not work with this addr from user space. In fact, it is mcr can not work with addr from user space, it would hung. Also, ion_vm_falut would happen twice. The first time is from __dabt_usr, when we access the mmaped buffer, it is fine. The second is from __davt_svc, it is caused by mcr, it is strange? ION malloc carveout heap addr = user mmap user access addr, ion_vm_fault (__dabt_usr), build page table, and vm_insert_page. dmac_map_area & dmac_flush_range with addr -> ion_vm_fault (__davt_svc) mcr hung. Not understand why ion_vm_fault happen twice, where page table has been build. Why mcr will hung with addr from user space. Besides, no problem with ION on 3.0, which do not use ion_vm_fault. Any suggestion? Thanks

12 years, 9 months

4
9
0 0

Patch to show debugfs output for every heap->id instead of heap->type

by Nishanth Peethambaran

Hi, ION debugfs currently shows/groups output based on type. But, it is possible to have multiple heaps of the same type - for CMA and carveout types. It is more useful to get usage information for individual heaps. - Nishanth Peethambaran >From fa819b42fb69321a8e5db260ba9fd8ce7a2f16d2 Mon Sep 17 00:00:00 2001 From: Nishanth Peethambaran <nishanth(a)broadcom.com> Date: Tue, 28 Aug 2012 07:57:37 +0530 Subject: [PATCH] gpu: ion: Update debugfs to show for each id Update the debugfs read of client and heap to show based on 'id' instead of 'type'. Multiple heaps of the same type can be present, but id is unique. --- drivers/gpu/ion/ion.c | 14 +++++++------- 1 files changed, 7 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/ion/ion.c b/drivers/gpu/ion/ion.c index 34c12df..65cedee 100644 --- a/drivers/gpu/ion/ion.c +++ b/drivers/gpu/ion/ion.c @@ -547,11 +547,11 @@ static int ion_debug_client_show(struct seq_file *s, void *unused) for (n = rb_first(&client->handles); n; n = rb_next(n)) { struct ion_handle *handle = rb_entry(n, struct ion_handle, node); - enum ion_heap_type type = handle->buffer->heap->type; + int id = handle->buffer->heap->id; - if (!names[type]) - names[type] = handle->buffer->heap->name; - sizes[type] += handle->buffer->size; + if (!names[id]) + names[id] = handle->buffer->heap->name; + sizes[id] += handle->buffer->size; } mutex_unlock(&client->lock); @@ -1121,7 +1121,7 @@ static const struct file_operations ion_fops = { }; static size_t ion_debug_heap_total(struct ion_client *client, - enum ion_heap_type type) + int id) { size_t size = 0; struct rb_node *n; @@ -1131,7 +1131,7 @@ static size_t ion_debug_heap_total(struct ion_client *client, struct ion_handle *handle = rb_entry(n, struct ion_handle, node); - if (handle->buffer->heap->type == type) + if (handle->buffer->heap->id == id) size += handle->buffer->size; } mutex_unlock(&client->lock); @@ -1149,7 +1149,7 @@ static int ion_debug_heap_show(struct seq_file *s, void *unused) for (n = rb_first(&dev->clients); n; n = rb_next(n)) { struct ion_client *client = rb_entry(n, struct ion_client, node); - size_t size = ion_debug_heap_total(client, heap->type); + size_t size = ion_debug_heap_total(client, heap->id); if (!size) continue; if (client->task) { -- 1.7.0.4

12 years, 9 months

1
0
0 0

sharing across process of ion/dma_buf created after fork

by Nishanth Peethambaran

How do we share ion buffers from user-space with other processes if they are exported/shared after fork? The ION_IOC_SHARE ioctl creates an fd for process-1. In 3.0 kernel, the ION_ION_IMPORT ioctl from process-2 calls ion_import_fd which calls fget(fd) which fails to find the file for the fd shared by process-1. In 3.4 kernel, dma_buf_get does the fget(fd) to get struct file which also fails for the same reason - fget searches in current->files. - Nishanth Peethambaran

12 years, 9 months

2
1
0 0

ioremap issue in ION carveout heap

by Haojian Zhuang

Hi all, I think that we have a memory mapping issue on ION carveout heap for v3.4+ kernel from android. The scenario is User app + kernel driver (cpu) + kernel driver (dma) that all these three clients will access memory. And the memory is cacheable. The .map_kernel() of carveout heap remaps the allocated memory buffer by ioremap(). In arm_ioremap(), we don't allow memory to be mapped. In order to make .map_kernel() working, we need to use memblock_alloc() & memblock_remove() to move the heap memory from system to reserved area. So the linear address of the memory buffer is removed from page table. And the new virtual address comes from .map_kernel() while kernel driver wants to access the buffer. But ION use dma_sync_sg_for_devices() to flush cache that means they're using linear address from page. So they're using the NOT-EXISTED virtual address that is removed by memblock_remove(). Solution #1. .map_kernel() only returns the linear address. And there's a limitation of this solution, the heap should be always lying in low memory. So we needn't use any ioremap() and memblock_remove() any more. Solution #2. Use vmap() in .map_kernel(). How do you think about these two solutions? Regards Haojian

12 years, 9 months

5
14
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

Linaro-mm-sig