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

[PATCH] mm: cma: fix alignment requirements for contiguous regions

by Marek Szyprowski

Contiguous Memory Allocator requires each of its regions to be aligned in such a way that it is possible to change migration type for all pageblocks holding it and then isolate page of largest possible order from the buddy allocator (which is MAX_ORDER-1). This patch relaxes alignment requirements by one order, because MAX_ORDER alignment is not really needed. Signed-off-by: Marek Szyprowski <m.szyprowski(a)samsung.com> CC: Michal Nazarewicz <mina86(a)mina86.com> --- drivers/base/dma-contiguous.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/drivers/base/dma-contiguous.c b/drivers/base/dma-contiguous.c index 78efb03..34d94c7 100644 --- a/drivers/base/dma-contiguous.c +++ b/drivers/base/dma-contiguous.c @@ -250,7 +250,7 @@ int __init dma_declare_contiguous(struct device *dev, unsigned long size, return -EINVAL; /* Sanitise input arguments */ - alignment = PAGE_SIZE << max(MAX_ORDER, pageblock_order); + alignment = PAGE_SIZE << max(MAX_ORDER - 1, pageblock_order); base = ALIGN(base, alignment); size = ALIGN(size, alignment); limit &= ~(alignment - 1); -- 1.7.1.569.g6f426

12 years, 9 months

2
1
0 0

[v3 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. v3: Provide a different path for IOMMU for more clean code. (Marek) atomic_pool is backed with struct page *pages[]. v2: Don't modify attrs(DMA_ATTR_NO_KERNEL_MAPPING) for atomic allocation. (Marek) Modify vzalloc (KyongHo, Minchan) 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 | 90 ++++++++++++++++++++++++++++++++++++++++----- 1 files changed, 80 insertions(+), 10 deletions(-) -- 1.7.5.4

12 years, 9 months

3
9
0 0

[RFC 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. *1: [ 8.321343] ------------[ cut here ]------------ [ 8.325971] kernel BUG at /home/hdoyu/mydroid-k340-cardhu/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: Refactor out to introduce __alloc_fill_pages ARM: dma-mapping: IOMMU allocates pages from pool with GFP_ATOMIC ARM: dma-mapping: Return cpu addr when dma_alloc(GFP_ATOMIC) ARM: dma-mapping: dma_{alloc,free}_coherent with empty attrs arch/arm/include/asm/dma-mapping.h | 21 ++++++++++-- arch/arm/mm/dma-mapping.c | 59 ++++++++++++++++++++++++++++-------- 2 files changed, 63 insertions(+), 17 deletions(-) -- 1.7.5.4

12 years, 9 months

4
15
0 0

[v2 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. v2: Don't modify attrs(DMA_ATTR_NO_KERNEL_MAPPING) for atomic allocation. (Marek) Skip vzalloc (KyongHo, Minchan) 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: Refactor out to introduce __in_atomic_pool ARM: dma-mapping: Use kzalloc() with GFP_ATOMIC ARM: dma-mapping: Refactor out to introduce __alloc_fill_pages ARM: dma-mapping: IOMMU allocates pages from atomic_pool with GFP_ATOMIC arch/arm/mm/dma-mapping.c | 102 ++++++++++++++++++++++++++++++++++----------- 1 files changed, 77 insertions(+), 25 deletions(-) -- 1.7.5.4

12 years, 9 months

3
6
0 0

[PATCHv6 0/2] ARM: replace custom consistent dma region with vmalloc

by Marek Szyprowski

Hello! This is yet another quick update on the patchset which replaces custom consistent dma regions usage in dma-mapping framework in favour of generic vmalloc areas created on demand for each allocation. The main purpose for this patchset is to remove 2MiB limit of dma coherent/writecombine allocations. This version addresses a few more cleanups pointed by Minchan Kim. This patch is based on vanilla v3.5 release. Best regards Marek Szyprowski Samsung Poland R&D Center Changelog: v6: - more cleanups of minor issues pointed by Minchan Kim, moved arm_dma_mmap() changes into separate patch v5: http://thread.gmane.org/gmane.linux.kernel.mm/83096 - fixed another minor issues pointed by Minchan Kim: added more comments here and there, changed pr_err() + stack_dump() to WARN(), added a fix for no-MMU systems v4: http://thread.gmane.org/gmane.linux.kernel.mm/80906 - replaced arch-independent VM_DMA flag with ARM-specific VM_ARM_DMA_CONSISTENT flag v3: http://thread.gmane.org/gmane.linux.kernel.mm/80028 - rebased onto v3.4-rc2: added support for IOMMU-aware implementation of dma-mapping calls, unified with CMA coherent dma pool - implemented changes requested by Minchan Kim: added more checks for vmarea->flags & VM_DMA, renamed some variables, removed obsole locks, squashed find_vm_area() exporting patch into the main redesign patch v2: http://thread.gmane.org/gmane.linux.kernel.mm/78563 - added support for atomic allocations (served from preallocated pool) - minor cleanup here and there - rebased onto v3.4-rc7 v1: http://thread.gmane.org/gmane.linux.kernel.mm/76703 - initial version Patch summary: Marek Szyprowski (2): mm: vmalloc: use const void * for caller argument ARM: dma-mapping: remove custom consistent dma region Documentation/kernel-parameters.txt | 2 +- arch/arm/include/asm/dma-mapping.h | 2 +- arch/arm/mm/dma-mapping.c | 486 ++++++++++++----------------------- arch/arm/mm/mm.h | 3 + include/linux/vmalloc.h | 9 +- mm/vmalloc.c | 28 ++- 6 files changed, 194 insertions(+), 336 deletions(-) -- 1.7.1.569.g6f426

12 years, 9 months

3
8
0 0

[PATCH] ARM: relax conditions required for enabling Contiguous Memory Allocator

by Marek Szyprowski

Contiguous Memory Allocator requires only paging and MMU enabled not particular CPU architectures, so there is no need for strict dependency on CPU type. This enables to use CMA on some older ARM v5 systems which also might need large contiguous blocks for the multimedia processing hw modules. Reported-by: Prabhakar Lad <prabhakar.csengg(a)gmail.com> Signed-off-by: Marek Szyprowski <m.szyprowski(a)samsung.com> --- arch/arm/Kconfig | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index e91c7cd..6ef75e2 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -6,7 +6,7 @@ config ARM select HAVE_DMA_API_DEBUG select HAVE_IDE if PCI || ISA || PCMCIA select HAVE_DMA_ATTRS - select HAVE_DMA_CONTIGUOUS if (CPU_V6 || CPU_V6K || CPU_V7) + select HAVE_DMA_CONTIGUOUS if MMU select HAVE_MEMBLOCK select RTC_LIB select SYS_SUPPORTS_APM_EMULATION -- 1.7.1.569.g6f426

12 years, 9 months

4
5
0 0

[PATCHv7 00/15] Integration of videobuf2 with dmabuf

by Tomasz Stanislawski

Hello everyone, This patchset adds support for DMABUF [2] importing to V4L2 stack. The support for DMABUF exporting was moved to separate patchset due to dependency on patches for DMA mapping redesign by Marek Szyprowski [4]. This patchset depends on new scatterlist constructor [5]. v7: - support for V4L2_MEMORY_DMABUF in v4l2-compact-ioctl32.c - cosmetic fixes to the documentation - added importing for vmalloc because vmap support in dmabuf for 3.5 was pull-requested - support for dmabuf importing for VIVI - resurrect allocation of dma-contig context - remove reference of alloc_ctx in dma-contig buffer - use sg_alloc_table_from_pages - fix DMA scatterlist calls to use orig_nents instead of nents - fix memleak in vb2_dc_sgt_foreach_page (use orig_nents instead of nents) v6: - fixed missing entry in v4l2_memory_names - fixed a bug occuring after get_user_pages failure - fixed a bug caused by using invalid vma for get_user_pages - prepare/finish no longer call dma_sync for dmabuf buffers v5: - removed change of importer/exporter behaviour - fixes vb2_dc_pages_to_sgt basing on Laurent's hints - changed pin/unpin words to lock/unlock in Doc v4: - rebased on mainline 3.4-rc2 - included missing importing support for s5p-fimc and s5p-tv - added patch for changing map/unmap for importers - fixes to Documentation part - coding style fixes - pairing {map/unmap}_dmabuf in vb2-core - fixing variable types and semantic of arguments in videobufb2-dma-contig.c v3: - rebased on mainline 3.4-rc1 - split 'code refactor' patch to multiple smaller patches - squashed fixes to Sumit's patches - patchset is no longer dependant on 'DMA mapping redesign' - separated path for handling IO and non-IO mappings - add documentation for DMABUF importing to V4L - removed all DMABUF exporter related code - removed usage of dma_get_pages extension v2: - extended VIDIOC_EXPBUF argument from integer memoffset to struct v4l2_exportbuffer - added patch that breaks DMABUF spec on (un)map_atachment callcacks but allows to work with existing implementation of DMABUF prime in DRM - all dma-contig code refactoring patches were squashed - bugfixes v1: List of changes since [1]. - support for DMA api extension dma_get_pages, the function is used to retrieve pages used to create DMA mapping. - small fixes/code cleanup to videobuf2 - added prepare and finish callbacks to vb2 allocators, it is used keep consistency between dma-cpu acess to the memory (by Marek Szyprowski) - support for exporting of DMABUF buffer in V4L2 and Videobuf2, originated from [3]. - support for dma-buf exporting in vb2-dma-contig allocator - support for DMABUF for s5p-tv and s5p-fimc (capture interface) drivers, originated from [3] - changed handling for userptr buffers (by Marek Szyprowski, Andrzej Pietrasiewicz) - let mmap method to use dma_mmap_writecombine call (by Marek Szyprowski) [1] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/4296… [2] https://lkml.org/lkml/2011/12/26/29 [3] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/3635… [4] http://thread.gmane.org/gmane.linux.kernel.cross-arch/12819 [5] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/47983 Laurent Pinchart (2): v4l: vb2-dma-contig: Shorten vb2_dma_contig prefix to vb2_dc v4l: vb2-dma-contig: Reorder functions Marek Szyprowski (2): v4l: vb2: add prepare/finish callbacks to allocators v4l: vb2-dma-contig: add prepare/finish to dma-contig allocator Sumit Semwal (4): v4l: Add DMABUF as a memory type v4l: vb2: add support for shared buffer (dma_buf) v4l: vb: remove warnings about MEMORY_DMABUF v4l: vb2-dma-contig: add support for dma_buf importing Tomasz Stanislawski (7): Documentation: media: description of DMABUF importing in V4L2 v4l: vb2-dma-contig: remove reference of alloc_ctx from a buffer v4l: vb2-dma-contig: add support for scatterlist in userptr mode v4l: vb2-vmalloc: add support for dmabuf importing v4l: vivi: support for dmabuf importing v4l: s5p-tv: mixer: support for dmabuf importing v4l: s5p-fimc: support for dmabuf importing Documentation/DocBook/media/v4l/compat.xml | 4 + Documentation/DocBook/media/v4l/io.xml | 179 ++++++++ .../DocBook/media/v4l/vidioc-create-bufs.xml | 3 +- Documentation/DocBook/media/v4l/vidioc-qbuf.xml | 15 + Documentation/DocBook/media/v4l/vidioc-reqbufs.xml | 47 +- drivers/media/video/Kconfig | 1 + drivers/media/video/s5p-fimc/Kconfig | 1 + drivers/media/video/s5p-fimc/fimc-capture.c | 2 +- drivers/media/video/s5p-tv/Kconfig | 1 + drivers/media/video/s5p-tv/mixer_video.c | 2 +- drivers/media/video/v4l2-compat-ioctl32.c | 16 + drivers/media/video/v4l2-ioctl.c | 1 + drivers/media/video/videobuf-core.c | 4 + drivers/media/video/videobuf2-core.c | 207 ++++++++- drivers/media/video/videobuf2-dma-contig.c | 470 +++++++++++++++++--- drivers/media/video/videobuf2-vmalloc.c | 56 +++ drivers/media/video/vivi.c | 2 +- include/linux/videodev2.h | 7 + include/media/videobuf2-core.h | 34 ++ 19 files changed, 963 insertions(+), 89 deletions(-) -- 1.7.9.5

12 years, 9 months

6
37
0 0

[PATCH v2 0/2] Enhance DMABUF with reference counting for exporter module

by Tomasz Stanislawski

Hello, This patchset adds reference counting for an exporter module to DMABUF framework. Moreover, it adds setup of an owner field for exporters in DRM subsystem. v1: Original v2: - split patch into DMABUF and DRM part - allow owner to be NULL Regards, Tomasz Stanislawski Tomasz Stanislawski (2): dma-buf: add reference counting for exporter module drm: set owner field to for all DMABUF exporters Documentation/dma-buf-sharing.txt | 3 ++- drivers/base/dma-buf.c | 9 ++++++++- drivers/gpu/drm/exynos/exynos_drm_dmabuf.c | 1 + drivers/gpu/drm/i915/i915_gem_dmabuf.c | 1 + drivers/gpu/drm/nouveau/nouveau_prime.c | 1 + drivers/gpu/drm/radeon/radeon_prime.c | 1 + drivers/staging/omapdrm/omap_gem_dmabuf.c | 1 + include/linux/dma-buf.h | 2 ++ 8 files changed, 17 insertions(+), 2 deletions(-) -- 1.7.9.5

12 years, 10 months

2
4
0 0

[PATCH 1/3] dma-fence: dma-buf synchronization (v7)

by Maarten Lankhorst

A dma-fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A dma-fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence. + dma_fence_signal() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. TODO maybe need some helper fxn for simple devices, like a display- only drm/kms device which simply wants to wait for exclusive fence to be signaled, and then attach a non-exclusive fence while scanout is in progress. The one pending on the fence can add an async callback: + dma_fence_add_callback() The callback can optionally be cancelled with remove_wait_queue() Or wait synchronously (optionally with timeout or interruptible): + dma_fence_wait() A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = dma_buf_get_fence(dmabuf); if (fence->ops == &bikeshed_fence_ops) { dma_buf *fence_buf; dma_bikeshed_fence_get_buf(fence, &fence_buf, &offset); ... tell the hw the memory location to wait on ... } else { /* fall-back to sw sync * / dma_fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. To facilitate other non-sw implementations, the enable_signaling callback can be used to keep track if a device not supporting hw sync is waiting on the fence, and in this case should arrange to call dma_fence_signal() at some point after the condition has changed, to notify other devices waiting on the fence. If there are no sw waiters, this can be skipped to avoid waking the CPU unnecessarily. The handler of the enable_signaling op should take a refcount until the fence is signaled, then release its ref. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw->hw signaling path (it can be handled same as sw->sw case), and therefore the fence->ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw->hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. v6: [ Maarten Lankhorst ] I removed dma_fence_cancel_callback and some comments about checking if fence fired or not. This is broken by design. waitqueue_active during destruction is now fatal, since the signaller should be holding a reference in enable_signalling until it signalled the fence. Pass the original dma_fence_cb along, and call __remove_wait in the dma_fence_callback handler, so that no cleanup needs to be performed. v7: [ Maarten Lankhorst ] Set cb->func and only enable sw signaling if fence wasn't signaled yet, for example for hardware fences that may choose to signal blindly. Signed-off-by: Maarten Lankhorst <maarten.lankhorst(a)canonical.com> --- drivers/base/Makefile | 2 drivers/base/dma-fence.c | 287 +++++++++++++++++++++++++++++++++++++++++++++ include/linux/dma-fence.h | 96 +++++++++++++++ 3 files changed, 384 insertions(+), 1 deletion(-) create mode 100644 drivers/base/dma-fence.c create mode 100644 include/linux/dma-fence.h diff --git a/drivers/base/Makefile b/drivers/base/Makefile index 5aa2d70..6e9f217 100644 --- a/drivers/base/Makefile +++ b/drivers/base/Makefile @@ -10,7 +10,7 @@ obj-$(CONFIG_CMA) += dma-contiguous.o obj-y += power/ obj-$(CONFIG_HAS_DMA) += dma-mapping.o obj-$(CONFIG_HAVE_GENERIC_DMA_COHERENT) += dma-coherent.o -obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o +obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o dma-fence.o obj-$(CONFIG_ISA) += isa.o obj-$(CONFIG_FW_LOADER) += firmware_class.o obj-$(CONFIG_NUMA) += node.o diff --git a/drivers/base/dma-fence.c b/drivers/base/dma-fence.c new file mode 100644 index 0000000..c280ee7 --- /dev/null +++ b/drivers/base/dma-fence.c @@ -0,0 +1,287 @@ +/* + * Fence mechanism for dma-buf to allow for asynchronous dma access + * + * Copyright (C) 2012 Texas Instruments + * Author: Rob Clark <rob.clark(a)linaro.org> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <linux/slab.h> +#include <linux/sched.h> +#include <linux/export.h> +#include <linux/dma-fence.h> + +/** + * dma_fence_signal - Signal a fence. + * + * @fence: The fence to signal + * + * All registered callbacks will be called directly (synchronously) and + * all blocked waters will be awoken. + * + * TODO: any value in adding a dma_fence_cancel(), for example to recov + * from hung gpu? It would behave like dma_fence_signal() but return + * an error to waiters and cb's to let them know that the condition they + * are waiting for will never happen. + */ +int dma_fence_signal(struct dma_fence *fence) +{ + unsigned long flags; + int ret = -EINVAL; + + if (WARN_ON(!fence)) + return -EINVAL; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!fence->signaled) { + fence->signaled = true; + __wake_up_locked_key(&fence->event_queue, TASK_NORMAL, + &fence->event_queue); + ret = 0; + } else WARN(1, "Already signaled"); + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_signal); + +static void release_fence(struct kref *kref) +{ + struct dma_fence *fence = + container_of(kref, struct dma_fence, refcount); + + BUG_ON(waitqueue_active(&fence->event_queue)); + + if (fence->ops->release) + fence->ops->release(fence); + + kfree(fence); +} + +/** + * dma_fence_put - Release a reference to the fence. + */ +void dma_fence_put(struct dma_fence *fence) +{ + WARN_ON(!fence); + kref_put(&fence->refcount, release_fence); +} +EXPORT_SYMBOL_GPL(dma_fence_put); + +/** + * dma_fence_get - Take a reference to the fence. + * + * In most cases this is used only internally by dma-fence. + */ +void dma_fence_get(struct dma_fence *fence) +{ + WARN_ON(!fence); + kref_get(&fence->refcount); +} +EXPORT_SYMBOL_GPL(dma_fence_get); + +static int check_signaling(struct dma_fence *fence) +{ + bool enable_signaling = false, signaled; + unsigned long flags; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + signaled = fence->signaled; + if (!signaled && !fence->needs_sw_signal) + enable_signaling = fence->needs_sw_signal = true; + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + if (enable_signaling) { + int ret; + + /* At this point, if enable_signaling returns any error + * a wakeup has to be performanced regardless. + * -ENOENT signals fence was already signaled. Any other error + * inidicates a catastrophic hardware error. + * + * If any hardware error occurs, nothing can be done against + * it, so it's treated like the fence was already signaled. + * No synchronization can be performed, so we have to assume + * the fence was already signaled. + */ + ret = fence->ops->enable_signaling(fence); + if (ret) { + signaled = true; + dma_fence_signal(fence); + } + } + + if (!signaled) + return 0; + else + return -ENOENT; +} + +int __dma_fence_wake_func(wait_queue_t *wait, unsigned mode, + int flags, void *key) +{ + struct dma_fence_cb *cb = + container_of(wait, struct dma_fence_cb, base); + + __remove_wait_queue(key, wait); + return cb->func(cb, wait->private); +} + +/** + * dma_fence_add_callback - Add a callback to be called when the fence + * is signaled. + * + * @fence: The fence to wait on + * @cb: The callback to register + * + * Any number of callbacks can be registered to a fence, but a callback + * can only be registered to once fence at a time. + * + * Note that the callback can be called from an atomic context. If + * fence is already signaled, this function will return -ENOENT (and + * *not* call the callback) + */ +int dma_fence_add_callback(struct dma_fence *fence, struct dma_fence_cb *cb, + dma_fence_func_t func, void *priv) +{ + unsigned long flags; + int ret; + + if (WARN_ON(!fence || !func)) + return -EINVAL; + + ret = check_signaling(fence); + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!ret && fence->signaled) + ret = -ENOENT; + + if (!ret) { + cb->base.flags = 0; + cb->base.func = __dma_fence_wake_func; + cb->base.private = priv; + cb->fence = fence; + cb->func = func; + __add_wait_queue(&fence->event_queue, &cb->base); + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_add_callback); + +/** + * dma_fence_wait - Wait for a fence to be signaled. + * + * @fence: The fence to wait on + * @interruptible: if true, do an interruptible wait + * @timeout: absolute time for timeout, in jiffies. + * + * Returns 0 on success, -EBUSY if a timeout occured, + * -ERESTARTSYS if the wait was interrupted by a signal. + */ +int dma_fence_wait(struct dma_fence *fence, bool interruptible, unsigned long timeout) +{ + unsigned long cur; + int ret; + + if (WARN_ON(!fence)) + return -EINVAL; + + cur = jiffies; + if (time_after_eq(cur, timeout)) + return -EBUSY; + + timeout -= cur; + + ret = check_signaling(fence); + if (ret == -ENOENT) + return 0; + else if (ret) + return ret; + + if (interruptible) + ret = wait_event_interruptible_timeout(fence->event_queue, + fence->signaled, + timeout); + else + ret = wait_event_timeout(fence->event_queue, + fence->signaled, timeout); + + if (ret > 0) + return 0; + else if (!ret) + return -EBUSY; + else + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_wait); + +/* + * Helpers intended to be used by the ops of the dma_fence implementation: + * + * NOTE: helpers and fxns intended to be used by other dma-fence + * implementations are not exported.. I'm not really sure if it makes + * sense to have a dma-fence implementation that is itself a module. + */ + +void __dma_fence_init(struct dma_fence *fence, struct dma_fence_ops *ops, void *priv) +{ + WARN_ON(!ops || !ops->enable_signaling); + + kref_init(&fence->refcount); + fence->ops = ops; + fence->priv = priv; + init_waitqueue_head(&fence->event_queue); +} +EXPORT_SYMBOL_GPL(__dma_fence_init); + +/* + * Pure sw implementation for dma-fence. The CPU always gets involved. + */ + +static int sw_enable_signaling(struct dma_fence *fence) +{ + /* + * pure sw, no irq's to enable, because the fence creator will + * always call dma_fence_signal() + */ + return 0; +} + +static struct dma_fence_ops sw_fence_ops = { + .enable_signaling = sw_enable_signaling, +}; + +/** + * dma_fence_create - Create a simple sw-only fence. + * + * This fence only supports signaling from/to CPU. Other implementations + * of dma-fence can be used to support hardware to hardware signaling, if + * supported by the hardware, and use the dma_fence_helper_* functions for + * compatibility with other devices that only support sw signaling. + */ +struct dma_fence *dma_fence_create(void) +{ + struct dma_fence *fence; + + fence = kzalloc(sizeof(struct dma_fence), GFP_KERNEL); + if (!fence) + return ERR_PTR(-ENOMEM); + + __dma_fence_init(fence, &sw_fence_ops, 0); + + return fence; +} +EXPORT_SYMBOL_GPL(dma_fence_create); diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h new file mode 100644 index 0000000..70d12c0 --- /dev/null +++ b/include/linux/dma-fence.h @@ -0,0 +1,96 @@ +/* + * Fence mechanism for dma-buf to allow for asynchronous dma access + * + * Copyright (C) 2012 Texas Instruments + * Author: Rob Clark <rob.clark(a)linaro.org> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#ifndef __DMA_FENCE_H__ +#define __DMA_FENCE_H__ + +#include <linux/err.h> +#include <linux/list.h> +#include <linux/wait.h> +#include <linux/list.h> +#include <linux/dma-buf.h> + +struct dma_fence; +struct dma_fence_ops; +struct dma_fence_cb; + +struct dma_fence { + struct kref refcount; + struct dma_fence_ops *ops; + wait_queue_head_t event_queue; + void *priv; + + /* has this fence been signaled yet? */ + bool signaled : 1; + + /* do we have one or more waiters or callbacks? */ + bool needs_sw_signal : 1; +}; + +typedef int (*dma_fence_func_t)(struct dma_fence_cb *cb, void *priv); + +struct dma_fence_cb { + wait_queue_t base; + dma_fence_func_t func; + struct dma_fence *fence; +}; + +struct dma_fence_ops { + /** + * For fence implementations that have the capability for hw->hw + * signaling, they can implement this op to enable the necessary + * irqs, or insert commands into cmdstream, etc. This is called + * in the first wait() or add_callback() path to let the fence + * implementation know that there is another driver waiting on + * the signal (ie. hw->sw case). + * + * A return value of -ENOENT will indicate that the fence has + * already passed. Any other errors will be treated as -ENOENT, + * and can happen because of hardware failure. + */ + int (*enable_signaling)(struct dma_fence *fence); + void (*release)(struct dma_fence *fence); +}; + +/* + * TODO does it make sense to be able to enable dma-fence without dma-buf, + * or visa versa? + */ +#ifdef CONFIG_DMA_SHARED_BUFFER + +/* create a basic (pure sw) fence: */ +struct dma_fence *dma_fence_create(void); + +/* intended to be used by other dma_fence implementations: */ +void __dma_fence_init(struct dma_fence *fence, + struct dma_fence_ops *ops, void *priv); + +void dma_fence_get(struct dma_fence *fence); +void dma_fence_put(struct dma_fence *fence); + +int dma_fence_signal(struct dma_fence *fence); +int dma_fence_wait(struct dma_fence *fence, bool interruptible, unsigned long timeout); +int dma_fence_add_callback(struct dma_fence *fence, struct dma_fence_cb *cb, + dma_fence_func_t func, void *priv); + +#else +// TODO +#endif /* CONFIG_DMA_SHARED_BUFFER */ + +#endif /* __DMA_FENCE_H__ */

12 years, 10 months

2
5
0 0

[GIT PULL] DMA-mapping fixups for v3.6-rc2

by Marek Szyprowski

Hi Linus, I would like to ask for pulling a set of fixup patches for ARM dma-mapping extensions merged in v3.6-rc1. The following changes since commit 0d7614f09c1ebdbaa1599a5aba7593f147bf96ee: Linux 3.6-rc1 (2012-08-02 16:38:10 -0700) with the top-most commit d9e0d149b5dcc2ef4688afc572b9906bcda941ef ARM: dma-mapping: fix incorrect freeing of atomic allocations are available in the git repository at: git://git.linaro.org/people/mszyprowski/linux-dma-mapping.git fixes-for-linus-for-3.6-rc2 Thanks! Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: Aaro Koskinen (2): ARM: dma-mapping: fix atomic allocation alignment ARM: dma-mapping: fix incorrect freeing of atomic allocations Chris Brand (1): ARM: mm: fix MMU mapping of CMA regions arch/arm/mm/dma-mapping.c | 12 ++++++------ 1 files changed, 6 insertions(+), 6 deletions(-)

12 years, 10 months

1
0
0 0

[PATCH] dma-buf: add reference counting for exporter module

by Tomasz Stanislawski

This patch adds reference counting on a module that exports dma-buf and implements its operations. This prevents the module from being unloaded while DMABUF file is in use. Signed-off-by: Tomasz Stanislawski <t.stanislaws(a)samsung.com> --- Documentation/dma-buf-sharing.txt | 3 ++- drivers/base/dma-buf.c | 10 +++++++++- drivers/gpu/drm/exynos/exynos_drm_dmabuf.c | 1 + drivers/gpu/drm/i915/i915_gem_dmabuf.c | 1 + drivers/gpu/drm/nouveau/nouveau_prime.c | 1 + drivers/gpu/drm/radeon/radeon_prime.c | 1 + drivers/staging/omapdrm/omap_gem_dmabuf.c | 1 + include/linux/dma-buf.h | 2 ++ 8 files changed, 18 insertions(+), 2 deletions(-) diff --git a/Documentation/dma-buf-sharing.txt b/Documentation/dma-buf-sharing.txt index ad86fb8..2613057 100644 --- a/Documentation/dma-buf-sharing.txt +++ b/Documentation/dma-buf-sharing.txt @@ -49,7 +49,8 @@ The dma_buf buffer sharing API usage contains the following steps: The buffer exporter announces its wish to export a buffer. In this, it connects its own private buffer data, provides implementation for operations that can be performed on the exported dma_buf, and flags for the file - associated with this buffer. + associated with this buffer. The operations structure has owner field. + You should initialize this to THIS_MODULE in most cases. Interface: struct dma_buf *dma_buf_export(void *priv, struct dma_buf_ops *ops, diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c index c30f3e1..d14b2f5 100644 --- a/drivers/base/dma-buf.c +++ b/drivers/base/dma-buf.c @@ -27,6 +27,7 @@ #include <linux/dma-buf.h> #include <linux/anon_inodes.h> #include <linux/export.h> +#include <linux/module.h> static inline int is_dma_buf_file(struct file *); @@ -40,6 +41,7 @@ static int dma_buf_release(struct inode *inode, struct file *file) dmabuf = file->private_data; dmabuf->ops->release(dmabuf); + module_put(dmabuf->ops->owner); kfree(dmabuf); return 0; } @@ -96,6 +98,7 @@ struct dma_buf *dma_buf_export(void *priv, const struct dma_buf_ops *ops, struct file *file; if (WARN_ON(!priv || !ops + || !ops->owner || !ops->map_dma_buf || !ops->unmap_dma_buf || !ops->release @@ -105,9 +108,14 @@ struct dma_buf *dma_buf_export(void *priv, const struct dma_buf_ops *ops, return ERR_PTR(-EINVAL); } + if (!try_module_get(ops->owner)) + return ERR_PTR(-ENOENT); + dmabuf = kzalloc(sizeof(struct dma_buf), GFP_KERNEL); - if (dmabuf == NULL) + if (dmabuf == NULL) { + module_put(ops->owner); return ERR_PTR(-ENOMEM); + } dmabuf->priv = priv; dmabuf->ops = ops; diff --git a/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c b/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c index 613bf8a..cf3bc6d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c @@ -164,6 +164,7 @@ static void exynos_gem_dmabuf_kunmap(struct dma_buf *dma_buf, } static struct dma_buf_ops exynos_dmabuf_ops = { + .owner = THIS_MODULE, .map_dma_buf = exynos_gem_map_dma_buf, .unmap_dma_buf = exynos_gem_unmap_dma_buf, .kmap = exynos_gem_dmabuf_kmap, diff --git a/drivers/gpu/drm/i915/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/i915_gem_dmabuf.c index aa308e1..07ff03b 100644 --- a/drivers/gpu/drm/i915/i915_gem_dmabuf.c +++ b/drivers/gpu/drm/i915/i915_gem_dmabuf.c @@ -152,6 +152,7 @@ static int i915_gem_dmabuf_mmap(struct dma_buf *dma_buf, struct vm_area_struct * } static const struct dma_buf_ops i915_dmabuf_ops = { + .owner = THIS_MODULE, .map_dma_buf = i915_gem_map_dma_buf, .unmap_dma_buf = i915_gem_unmap_dma_buf, .release = i915_gem_dmabuf_release, diff --git a/drivers/gpu/drm/nouveau/nouveau_prime.c b/drivers/gpu/drm/nouveau/nouveau_prime.c index a25cf2c..8605033 100644 --- a/drivers/gpu/drm/nouveau/nouveau_prime.c +++ b/drivers/gpu/drm/nouveau/nouveau_prime.c @@ -127,6 +127,7 @@ static void nouveau_gem_prime_vunmap(struct dma_buf *dma_buf, void *vaddr) } static const struct dma_buf_ops nouveau_dmabuf_ops = { + .owner = THIS_MODULE, .map_dma_buf = nouveau_gem_map_dma_buf, .unmap_dma_buf = nouveau_gem_unmap_dma_buf, .release = nouveau_gem_dmabuf_release, diff --git a/drivers/gpu/drm/radeon/radeon_prime.c b/drivers/gpu/drm/radeon/radeon_prime.c index 6bef46a..4061fd3 100644 --- a/drivers/gpu/drm/radeon/radeon_prime.c +++ b/drivers/gpu/drm/radeon/radeon_prime.c @@ -127,6 +127,7 @@ static void radeon_gem_prime_vunmap(struct dma_buf *dma_buf, void *vaddr) mutex_unlock(&dev->struct_mutex); } const static struct dma_buf_ops radeon_dmabuf_ops = { + .owner = THIS_MODULE, .map_dma_buf = radeon_gem_map_dma_buf, .unmap_dma_buf = radeon_gem_unmap_dma_buf, .release = radeon_gem_dmabuf_release, diff --git a/drivers/staging/omapdrm/omap_gem_dmabuf.c b/drivers/staging/omapdrm/omap_gem_dmabuf.c index 42728e0..6a4dd67 100644 --- a/drivers/staging/omapdrm/omap_gem_dmabuf.c +++ b/drivers/staging/omapdrm/omap_gem_dmabuf.c @@ -179,6 +179,7 @@ out_unlock: } struct dma_buf_ops omap_dmabuf_ops = { + .owner = THIS_MODULE, .map_dma_buf = omap_gem_map_dma_buf, .unmap_dma_buf = omap_gem_unmap_dma_buf, .release = omap_gem_dmabuf_release, diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h index eb48f38..22953de 100644 --- a/include/linux/dma-buf.h +++ b/include/linux/dma-buf.h @@ -37,6 +37,7 @@ struct dma_buf_attachment; /** * struct dma_buf_ops - operations possible on struct dma_buf + * @owner: the module that implements dma_buf operations * @attach: [optional] allows different devices to 'attach' themselves to the * given buffer. It might return -EBUSY to signal that backing storage * is already allocated and incompatible with the requirements @@ -70,6 +71,7 @@ struct dma_buf_attachment; * @vunmap: [optional] unmaps a vmap from the buffer */ struct dma_buf_ops { + struct module *owner; int (*attach)(struct dma_buf *, struct device *, struct dma_buf_attachment *); -- 1.7.9.5

12 years, 10 months

4
4
0 0

[PATCHv2 0/9] Support for dmabuf exporting for videobuf2

by Tomasz Stanislawski

Hello everyone, The patches adds support for DMABUF exporting to V4L2 stack. The latest support for DMABUF importing was posted in [1]. The exporter part is dependant on DMA mapping redesign [2] which is expected to be merged into the mainline. Therefore it is posted as a separate patchset. Moreover some patches depends on vmap extension for DMABUF by Dave Airlie [3] and sg_alloc_table_from_pages function [4]. The last patch 'v4l: vb2-dma-contig: use dma_get_sgtable' depends on dma_get_sgtable extension to DMA api [5]. The tree with all the patches and extensions is available at: repo: git://git.infradead.org/users/kmpark/linux-2.6-samsung branch: media-for3.5-vb2-dmabuf-v7 Changelog: v2: - add documentation for DMABUF exporting - squashed 'let mmap method to use dma_mmap_coherent call' with 'remove vb2_mmap_pfn_range function' - move setup of scatterlist for MMAP buffers from alloc to DMABUF export code - use locking to serialize map/unmap of DMABUF attachments - squash vmap/kmap, setup of sg lists, allocation in attachments into dma-contig exporter patch - fix occasional failure of follow_pfn trick by using init_mm in artificial VMA - add support for exporting in s5p-mfc driver - drop all code that duplicates sg_alloc_table_from_pages - introduce usage of dma_get_sgtable as generic solution to follow_pfn trick v1: - updated setup of VIDIOC_EXPBUF ioctl - doc updates - introduced workaround to avoid using dma_get_pages, - removed caching of exported dmabuf to avoid existence of circular reference between dmabuf and vb2_dc_buf or resource leakage - removed all 'change behaviour' patches - inital support for exporting in s5p-mfs driver - removal of vb2_mmap_pfn_range that is no longer used - use sg_alloc_table_from_pages instead of creating sglist in vb2_dc code - move attachment allocation to exporter's attach callback v0: RFC - initial version [1] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/49438 [2] http://thread.gmane.org/gmane.linux.kernel.cross-arch/14098 [3] http://permalink.gmane.org/gmane.comp.video.dri.devel/69302 [4] This patchset is rebased on 3.4-rc1 plus the following patchsets: [5] http://www.spinics.net/lists/linux-arch/msg18282.html Marek Szyprowski (1): v4l: vb2-dma-contig: let mmap method to use dma_mmap_coherent call Tomasz Stanislawski (8): Documentation: media: description of DMABUF exporting in V4L2 v4l: add buffer exporting via dmabuf v4l: vb2: add buffer exporting via dmabuf v4l: vb2-dma-contig: add support for DMABUF exporting v4l: s5p-fimc: support for dmabuf exporting v4l: s5p-tv: mixer: support for dmabuf exporting v4l: s5p-mfc: support for dmabuf exporting v4l: vb2-dma-contig: use dma_get_sgtable Documentation/DocBook/media/v4l/compat.xml | 3 + Documentation/DocBook/media/v4l/io.xml | 3 + Documentation/DocBook/media/v4l/v4l2.xml | 1 + Documentation/DocBook/media/v4l/vidioc-expbuf.xml | 223 ++++++++++++++++++++ drivers/media/video/s5p-fimc/fimc-capture.c | 9 + drivers/media/video/s5p-mfc/s5p_mfc_dec.c | 18 ++ drivers/media/video/s5p-mfc/s5p_mfc_enc.c | 18 ++ drivers/media/video/s5p-tv/mixer_video.c | 10 + drivers/media/video/v4l2-compat-ioctl32.c | 1 + drivers/media/video/v4l2-dev.c | 1 + drivers/media/video/v4l2-ioctl.c | 6 + drivers/media/video/videobuf2-core.c | 67 ++++++ drivers/media/video/videobuf2-dma-contig.c | 224 ++++++++++++++++++++- drivers/media/video/videobuf2-memops.c | 40 ---- include/linux/videodev2.h | 26 +++ include/media/v4l2-ioctl.h | 2 + include/media/videobuf2-core.h | 2 + include/media/videobuf2-memops.h | 5 - 18 files changed, 612 insertions(+), 47 deletions(-) create mode 100644 Documentation/DocBook/media/v4l/vidioc-expbuf.xml -- 1.7.9.5

12 years, 10 months

7
33
0 0

[PATCH 0/2] dma-parms and helpers for dma-buf

by Rob Clark

From: Rob Clark <rob(a)ti.com> Re-sending first patch, with a wider audience. Apparently I didn't spam enough inboxes the first time. And, at Daniel Vetter's suggestion, adding some helper functions in dma-buf to get the most restrictive parameters of all the attached devices. Rob Clark (2): device: add dma_params->max_segment_count dma-buf: add helpers for attacher dma-parms drivers/base/dma-buf.c | 63 +++++++++++++++++++++++++++++++++++++++++++ include/linux/device.h | 1 + include/linux/dma-buf.h | 19 +++++++++++++ include/linux/dma-mapping.h | 16 +++++++++++ 4 files changed, 99 insertions(+) -- 1.7.9.5

12 years, 10 months

5
8
0 0

[RFC PATCH 1/3] dma-fence: dma-buf synchronization (v5)

by Maarten Lankhorst

A dma-fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A dma-fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence. + dma_fence_signal() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. TODO maybe need some helper fxn for simple devices, like a display- only drm/kms device which simply wants to wait for exclusive fence to be signaled, and then attach a non-exclusive fence while scanout is in progress. The one pending on the fence can add an async callback (and optionally cancel it.. for example, to recover from GPU hangs): + dma_fence_add_callback() + dma_fence_cancel_callback() Or wait synchronously (optionally with timeout or from atomic context): + dma_fence_wait() A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = dma_buf_get_fence(dmabuf); if (fence->ops == &bikeshed_fence_ops) { dma_buf *fence_buf; dma_bikeshed_fence_get_buf(fence, &fence_buf, &offset); ... tell the hw the memory location to wait on ... } else { /* fall-back to sw sync * / dma_fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. To facilitate other non-sw implementations, the enable_signaling callback can be used to keep track if a device not supporting hw sync is waiting on the fence, and in this case should arrange to call dma_fence_signal() at some point after the condition has changed, to notify other devices waiting on the fence. If there are no sw waiters, this can be skipped to avoid waking the CPU unnecessarily. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw->hw signaling path (it can be handled same as sw->sw case), and therefore the fence->ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw->hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. --- drivers/base/Makefile | 2 drivers/base/dma-fence.c | 317 +++++++++++++++++++++++++++++++++++++++++++++ include/linux/dma-fence.h | 123 +++++++++++++++++ 3 files changed, 441 insertions(+), 1 deletion(-) create mode 100644 drivers/base/dma-fence.c create mode 100644 include/linux/dma-fence.h diff --git a/drivers/base/Makefile b/drivers/base/Makefile index 5aa2d70..6e9f217 100644 --- a/drivers/base/Makefile +++ b/drivers/base/Makefile @@ -10,7 +10,7 @@ obj-$(CONFIG_CMA) += dma-contiguous.o obj-y += power/ obj-$(CONFIG_HAS_DMA) += dma-mapping.o obj-$(CONFIG_HAVE_GENERIC_DMA_COHERENT) += dma-coherent.o -obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o +obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o dma-fence.o obj-$(CONFIG_ISA) += isa.o obj-$(CONFIG_FW_LOADER) += firmware_class.o obj-$(CONFIG_NUMA) += node.o diff --git a/drivers/base/dma-fence.c b/drivers/base/dma-fence.c new file mode 100644 index 0000000..6798dc4 --- /dev/null +++ b/drivers/base/dma-fence.c @@ -0,0 +1,317 @@ +/* + * Fence mechanism for dma-buf to allow for asynchronous dma access + * + * Copyright (C) 2012 Texas Instruments + * Author: Rob Clark <rob.clark(a)linaro.org> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <linux/slab.h> +#include <linux/sched.h> +#include <linux/export.h> +#include <linux/dma-fence.h> + +/** + * dma_fence_signal - Signal a fence. + * + * @fence: The fence to signal + * + * All registered callbacks will be called directly (synchronously) and + * all blocked waters will be awoken. + * + * TODO: any value in adding a dma_fence_cancel(), for example to recov + * from hung gpu? It would behave like dma_fence_signal() but return + * an error to waiters and cb's to let them know that the condition they + * are waiting for will never happen. + */ +int dma_fence_signal(struct dma_fence *fence) +{ + unsigned long flags; + int ret = -EINVAL; + + if (WARN_ON(!fence)) + return -EINVAL; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!fence->signaled) { + fence->signaled = true; + wake_up_all_locked(&fence->event_queue); + ret = 0; + } else WARN(1, "Already signaled"); + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_signal); + +static void release_fence(struct kref *kref) +{ + struct dma_fence *fence = + container_of(kref, struct dma_fence, refcount); + + WARN_ON(waitqueue_active(&fence->event_queue)); + if (fence->ops->release) + fence->ops->release(fence); + + kfree(fence); +} + +/** + * dma_fence_put - Release a reference to the fence. + */ +void dma_fence_put(struct dma_fence *fence) +{ + WARN_ON(!fence); + kref_put(&fence->refcount, release_fence); +} +EXPORT_SYMBOL_GPL(dma_fence_put); + +/** + * dma_fence_get - Take a reference to the fence. + * + * In most cases this is used only internally by dma-fence. + */ +void dma_fence_get(struct dma_fence *fence) +{ + WARN_ON(!fence); + kref_get(&fence->refcount); +} +EXPORT_SYMBOL_GPL(dma_fence_get); + +static int check_signaling(struct dma_fence *fence) +{ + bool enable_signaling = false, signaled; + unsigned long flags; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!fence->needs_sw_signal) + enable_signaling = fence->needs_sw_signal = true; + signaled = fence->signaled; + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + if (enable_signaling) { + int ret; + + /* At this point, if enable_signaling returns any error + * a wakeup has to be performanced regardless. + * -ENOENT signals fence was already signaled. Any other error + * inidicates a catastrophic hardware error. + * + * If any hardware error occurs, nothing can be done against + * it, so it's treated like the fence was already signaled. + * No synchronization can be performed, so we have to assume + * the fence was already signaled. + */ + ret = fence->ops->enable_signaling(fence); + if (ret) { + signaled = true; + dma_fence_signal(fence); + } + } + + if (!signaled) + return 0; + else + return -ENOENT; +} + +/** + * dma_fence_add_callback - Add a callback to be called when the fence + * is signaled. + * + * @fence: The fence to wait on + * @cb: The callback to register + * + * Any number of callbacks can be registered to a fence, but a callback + * can only be registered to once fence at a time. + * + * Note that the callback can be called from an atomic context. If + * fence is already signaled, this function will return -ENOENT (and + * *not* call the callback) + */ +int dma_fence_add_callback(struct dma_fence *fence, + struct dma_fence_cb *cb) +{ + unsigned long flags; + int ret; + + if (WARN_ON(!fence || !cb)) + return -EINVAL; + + ret = check_signaling(fence); + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!ret && !fence->signaled) { + cb->fence = fence; + __add_wait_queue(&fence->event_queue, &cb->base); + ret = 0; + } else if (!ret) + ret = -ENOENT; + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_add_callback); + +/** + * dma_fence_cancel_callback - Remove a previously registered callback. + * + * @cb: The callback to unregister + * + * The callback will not be called after this function returns, but could + * be called before this function returns. + */ +int dma_fence_cancel_callback(struct dma_fence_cb *cb) +{ + struct dma_fence *fence; + unsigned long flags; + int ret = -EINVAL; + + if (WARN_ON(!cb)) + return -EINVAL; + + fence = cb->fence; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (fence) { + __remove_wait_queue(&fence->event_queue, &cb->base); + cb->fence = NULL; + ret = 0; + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_cancel_callback); + +/** + * dma_fence_wait - Wait for a fence to be signaled. + * + * @fence: The fence to wait on + * @interruptible: if true, do an interruptible wait + * @timeout: absolute time for timeout, in jiffies. + * + * Returns 0 on success, -EBUSY if a timeout occured, + * -ERESTARTSYS if the wait was interrupted by a signal. + */ +int dma_fence_wait(struct dma_fence *fence, bool interruptible, unsigned long timeout) +{ + unsigned long cur; + int ret; + + if (WARN_ON(!fence)) + return -EINVAL; + + cur = jiffies; + if (time_after_eq(cur, timeout)) + return -EBUSY; + + timeout -= cur; + + ret = check_signaling(fence); + if (ret == -ENOENT) + return 0; + else if (ret) + return ret; + + if (interruptible) + ret = wait_event_interruptible_timeout(fence->event_queue, + fence->signaled, + timeout); + else + ret = wait_event_timeout(fence->event_queue, + fence->signaled, timeout); + + WARN(1, "wait_event_timeout(%u) returns %i", interruptible, ret); + if (ret > 0) + return 0; + else if (!ret) + return -EBUSY; + else + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_wait); + +int __dma_fence_wake_func(wait_queue_t *wait, unsigned mode, + int flags, void *key) +{ + struct dma_fence_cb *cb = + container_of(wait, struct dma_fence_cb, base); + struct dma_fence *fence = cb->fence; + int ret; + + ret = cb->func(cb, fence); + cb->fence = NULL; + + return ret; +} +EXPORT_SYMBOL_GPL(__dma_fence_wake_func); + +/* + * Helpers intended to be used by the ops of the dma_fence implementation: + * + * NOTE: helpers and fxns intended to be used by other dma-fence + * implementations are not exported.. I'm not really sure if it makes + * sense to have a dma-fence implementation that is itself a module. + */ + +void __dma_fence_init(struct dma_fence *fence, struct dma_fence_ops *ops, void *priv) +{ + WARN_ON(!ops || !ops->enable_signaling); + + kref_init(&fence->refcount); + fence->ops = ops; + fence->priv = priv; + init_waitqueue_head(&fence->event_queue); +} +EXPORT_SYMBOL_GPL(__dma_fence_init); + +/* + * Pure sw implementation for dma-fence. The CPU always gets involved. + */ + +static int sw_enable_signaling(struct dma_fence *fence) +{ + /* + * pure sw, no irq's to enable, because the fence creator will + * always call dma_fence_signal() + */ + return 0; +} + +static struct dma_fence_ops sw_fence_ops = { + .enable_signaling = sw_enable_signaling, +}; + +/** + * dma_fence_create - Create a simple sw-only fence. + * + * This fence only supports signaling from/to CPU. Other implementations + * of dma-fence can be used to support hardware to hardware signaling, if + * supported by the hardware, and use the dma_fence_helper_* functions for + * compatibility with other devices that only support sw signaling. + */ +struct dma_fence *dma_fence_create(void) +{ + struct dma_fence *fence; + + fence = kzalloc(sizeof(struct dma_fence), GFP_KERNEL); + if (!fence) + return ERR_PTR(-ENOMEM); + + __dma_fence_init(fence, &sw_fence_ops, 0); + + return fence; +} +EXPORT_SYMBOL_GPL(dma_fence_create); diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h new file mode 100644 index 0000000..648f136 --- /dev/null +++ b/include/linux/dma-fence.h @@ -0,0 +1,123 @@ +/* + * Fence mechanism for dma-buf to allow for asynchronous dma access + * + * Copyright (C) 2012 Texas Instruments + * Author: Rob Clark <rob.clark(a)linaro.org> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#ifndef __DMA_FENCE_H__ +#define __DMA_FENCE_H__ + +#include <linux/err.h> +#include <linux/list.h> +#include <linux/wait.h> +#include <linux/list.h> +#include <linux/dma-buf.h> + +struct dma_fence; +struct dma_fence_ops; +struct dma_fence_cb; + +struct dma_fence { + struct kref refcount; + struct dma_fence_ops *ops; + wait_queue_head_t event_queue; + void *priv; + + /* has this fence been signaled yet? */ + bool signaled : 1; + + /* do we have one or more waiters or callbacks? */ + bool needs_sw_signal : 1; +}; + +typedef int (*dma_fence_func_t)(struct dma_fence_cb *cb, + struct dma_fence *fence); + +struct dma_fence_cb { + wait_queue_t base; + dma_fence_func_t func; + + /* + * This is initialized when the cb is added, and NULL'd when it + * is canceled or expired, so can be used to for error checking + * if the cb is already pending. A dma_fence_cb can be pending + * on at most one fence at a time. + */ + struct dma_fence *fence; +}; + +struct dma_fence_ops { + /** + * For fence implementations that have the capability for hw->hw + * signaling, they can implement this op to enable the necessary + * irqs, or insert commands into cmdstream, etc. This is called + * in the first wait() or add_callback() path to let the fence + * implementation know that there is another driver waiting on + * the signal (ie. hw->sw case). + * + * A return value of -ENOENT will indicate that the fence has + * already passed. + */ + int (*enable_signaling)(struct dma_fence *fence); + void (*release)(struct dma_fence *fence); +}; + +int __dma_fence_wake_func(wait_queue_t *wait, unsigned mode, + int flags, void *key); + +#define DMA_FENCE_CB_INITIALIZER(cb_func) { \ + .base = { .func = __dma_fence_wake_func }, \ + .func = (cb_func), \ + } + +#define DECLARE_DMA_FENCE_CB(name, cb_func) \ + struct dma_fence_cb name = DMA_FENCE_CB_INITIALIZER(cb_func) + + +/* + * TODO does it make sense to be able to enable dma-fence without dma-buf, + * or visa versa? + */ +#ifdef CONFIG_DMA_SHARED_BUFFER + +/* create a basic (pure sw) fence: */ +struct dma_fence *dma_fence_create(void); + +/* intended to be used by other dma_fence implementations: */ +void __dma_fence_init(struct dma_fence *fence, struct dma_fence_ops *ops, void *priv); + +void dma_fence_get(struct dma_fence *fence); +void dma_fence_put(struct dma_fence *fence); +int dma_fence_signal(struct dma_fence *fence); + +int dma_fence_add_callback(struct dma_fence *fence, + struct dma_fence_cb *cb); +int dma_fence_cancel_callback(struct dma_fence_cb *cb); +int dma_fence_wait(struct dma_fence *fence, bool interruptible, unsigned long timeout); + +/* helpers intended to be used by the ops of the dma_fence implementation: */ +int dma_fence_helper_signal(struct dma_fence *fence); +int dma_fence_helper_add_callback(struct dma_fence *fence, + struct dma_fence_cb *cb); +int dma_fence_helper_cancel_callback(struct dma_fence_cb *cb); +int dma_fence_helper_wait(struct dma_fence *fence, bool interruptible, + long timeout); + +#else +// TODO +#endif /* CONFIG_DMA_SHARED_BUFFER */ + +#endif /* __DMA_FENCE_H__ */

12 years, 10 months

2
4
0 0

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

by Marek Szyprowski

Hi Linus, I would like to ask for pulling another set of DMA-mapping framework updates for v3.6. The following changes since commit 28a33cbc24e4256c143dce96c7d93bf423229f92: Linux 3.5 (2012-07-21 13:58:29 -0700) with the top-most commit 97ef952a20853fad72087a53fa556fbec45edd8f ARM: dma-mapping: add support for DMA_ATTR_SKIP_CPU_SYNC attribute are available in the git repository at: git://git.linaro.org/people/mszyprowski/linux-dma-mapping.git for-linus-for-3.6-rc1 Those patches are continuation of my earlier works merged in commits 58bca4a8fa90 and d484864dd96e1. They contains extensions to DMA-mapping framework to remove limitation of the current ARM implementation (like limited total size of DMA coherent/write combine buffers), improve performance of buffer sharing between devices (attributes to skip cpu cache operations or creation of additional kernel mapping for some specific use cases) as well as some unification of the common code for dma_mmap_attrs()/dma_mmap_coherent() functions. All extensions have been implemented and tested for ARM architecture. Thanks! Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: Marek Szyprowski (11): mm: vmalloc: use const void * for caller argument ARM: dma-mapping: remove custom consistent dma region ARM: dma-mapping: add more sanity checks in arm_dma_mmap() ARM: dma-mapping: fix error path for memory allocation failure common: dma-mapping: add support for generic dma_mmap_* calls common: DMA-mapping: add DMA_ATTR_NO_KERNEL_MAPPING attribute ARM: dma-mapping: add support for DMA_ATTR_NO_KERNEL_MAPPING attribute common: dma-mapping: introduce dma_get_sgtable() function ARM: dma-mapping: add support for dma_get_sgtable() common: DMA-mapping: add DMA_ATTR_SKIP_CPU_SYNC attribute ARM: dma-mapping: add support for DMA_ATTR_SKIP_CPU_SYNC attribute Tomasz Stanislawski (1): scatterlist: add sg_alloc_table_from_pages function Documentation/DMA-attributes.txt | 42 +++ Documentation/kernel-parameters.txt | 2 +- arch/arm/common/dmabounce.c | 1 + arch/arm/include/asm/dma-mapping.h | 24 +- arch/arm/mm/dma-mapping.c | 561 ++++++++++++------------------ arch/arm/mm/mm.h | 3 + arch/powerpc/include/asm/dma-mapping.h | 8 +- arch/powerpc/kernel/dma-iommu.c | 1 + arch/powerpc/kernel/dma-swiotlb.c | 1 + arch/powerpc/kernel/dma.c | 36 +- arch/powerpc/kernel/vio.c | 1 + drivers/base/dma-mapping.c | 49 +++ include/asm-generic/dma-coherent.h | 1 + include/asm-generic/dma-mapping-common.h | 55 +++ include/linux/dma-attrs.h | 2 + include/linux/dma-mapping.h | 3 + include/linux/scatterlist.h | 4 + include/linux/vmalloc.h | 9 +- lib/scatterlist.c | 64 ++++ mm/vmalloc.c | 28 +- 20 files changed, 505 insertions(+), 390 deletions(-)

12 years, 10 months

1
0
0 0

[PATCH] ARM: dma-mapping: add more sanity checks in arm_dma_mmap()

by Marek Szyprowski

Add some sanity checks and forbid mmaping of buffers into vma areas larger than allocated dma buffer. Signed-off-by: Marek Szyprowski <m.szyprowski(a)samsung.com> --- arch/arm/mm/dma-mapping.c | 14 ++++++++++---- 1 files changed, 10 insertions(+), 4 deletions(-) diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c index f906d5f..a2881c9 100644 --- a/arch/arm/mm/dma-mapping.c +++ b/arch/arm/mm/dma-mapping.c @@ -611,16 +611,22 @@ int arm_dma_mmap(struct device *dev, struct vm_area_struct *vma, { int ret = -ENXIO; #ifdef CONFIG_MMU + unsigned long nr_vma_pages = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; + unsigned long nr_pages = PAGE_ALIGN(size) >> PAGE_SHIFT; unsigned long pfn = dma_to_pfn(dev, dma_addr); + unsigned long off = vma->vm_pgoff; + vma->vm_page_prot = __get_dma_pgprot(attrs, vma->vm_page_prot); if (dma_mmap_from_coherent(dev, vma, cpu_addr, size, &ret)) return ret; - ret = remap_pfn_range(vma, vma->vm_start, - pfn + vma->vm_pgoff, - vma->vm_end - vma->vm_start, - vma->vm_page_prot); + if (off < nr_pages && nr_vma_pages <= (nr_pages - off)) { + ret = remap_pfn_range(vma, vma->vm_start, + pfn + off, + vma->vm_end - vma->vm_start, + vma->vm_page_prot); + } #endif /* CONFIG_MMU */ return ret; -- 1.7.1.569.g6f426

12 years, 10 months

1
0
0 0

[PATCHv5 0/2] ARM: replace custom consistent dma region with vmalloc

by Marek Szyprowski

Hello! This is yet another quick update on the patchset which replaces custom consistent dma regions usage in dma-mapping framework in favour of generic vmalloc areas created on demand for each allocation. The main purpose for this patchset is to remove 2MiB limit of dma coherent/writecombine allocations. This version addresses a few minor issues pointed by Minchan Kim. This patch is based on vanilla v3.5 release. Best regards Marek Szyprowski Samsung Poland R&D Center Changelog: v5: - fixed another minor issues pointed by Minchan Kim: added more comments here and there, changed pr_err() + stack_dump() to WARN(), added a fix for no-MMU systems v4: http://thread.gmane.org/gmane.linux.kernel.mm/80906 - replaced arch-independent VM_DMA flag with ARM-specific VM_ARM_DMA_CONSISTENT flag v3: http://thread.gmane.org/gmane.linux.kernel.mm/80028 - rebased onto v3.4-rc2: added support for IOMMU-aware implementation of dma-mapping calls, unified with CMA coherent dma pool - implemented changes requested by Minchan Kim: added more checks for vmarea->flags & VM_DMA, renamed some variables, removed obsole locks, squashed find_vm_area() exporting patch into the main redesign patch v2: http://thread.gmane.org/gmane.linux.kernel.mm/78563 - added support for atomic allocations (served from preallocated pool) - minor cleanup here and there - rebased onto v3.4-rc7 v1: http://thread.gmane.org/gmane.linux.kernel.mm/76703 - initial version Patch summary: Marek Szyprowski (2): mm: vmalloc: use const void * for caller argument ARM: dma-mapping: remove custom consistent dma region Documentation/kernel-parameters.txt | 2 +- arch/arm/include/asm/dma-mapping.h | 2 +- arch/arm/mm/dma-mapping.c | 511 +++++++++++++---------------------- arch/arm/mm/mm.h | 3 + include/linux/vmalloc.h | 9 +- mm/vmalloc.c | 28 ++- 6 files changed, 210 insertions(+), 345 deletions(-) -- 1.7.1.569.g6f426

12 years, 10 months

2
3
0 0

[PATCH] ARM: DMA-mapping: fix error path for memory allocation failure

by Marek Szyprowski

This patch fixes incorrect check in error path. When the allocation of first page fails, the kernel ops appears due to accessing -1 element of the pages array. Reported-by: Sylwester Nawrocki <s.nawrocki(a)samsung.com> Signed-off-by: Marek Szyprowski <m.szyprowski(a)samsung.com> --- arch/arm/mm/dma-mapping.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c index 7a1d224..3096ed2 100644 --- a/arch/arm/mm/dma-mapping.c +++ b/arch/arm/mm/dma-mapping.c @@ -977,7 +977,7 @@ static struct page **__iommu_alloc_buffer(struct device *dev, size_t size, gfp_t return pages; error: - while (--i) + while (i--) if (pages[i]) __free_pages(pages[i], 0); if (array_size <= PAGE_SIZE) -- 1.7.1.569.g6f426

12 years, 10 months

1
0
0 0

[RFc] Map CMA pages as cached

by Laura Abbott

Current APIs only support allocating CMA pages as either coherent or writecombine. This seems to miss support for cached pages completely. The following patch seems to be the first obvious solution. More generally though, what should be the strategy for remapping with other memory types? Add a new function call each time? Thanks, Laura

12 years, 10 months

4
7
0 0

[PATCH] device: add dma_params->max_segment_count

by Rob Clark

From: Rob Clark <rob(a)ti.com> For devices which have constraints about maximum number of segments in an sglist. For example, a device which could only deal with contiguous buffers would set max_segment_count to 1. The initial motivation is for devices sharing buffers via dma-buf, to allow the buffer exporter to know the constraints of other devices which have attached to the buffer. The dma_mask and fields in 'struct device_dma_parameters' tell the exporter everything else that is needed, except whether the importer has constraints about maximum number of segments. Signed-off-by: Rob Clark <rob(a)ti.com> --- include/linux/device.h | 1 + include/linux/dma-mapping.h | 16 ++++++++++++++++ 2 files changed, 17 insertions(+) diff --git a/include/linux/device.h b/include/linux/device.h index 161d962..3813735 100644 --- a/include/linux/device.h +++ b/include/linux/device.h @@ -568,6 +568,7 @@ struct device_dma_parameters { * sg limitations. */ unsigned int max_segment_size; + unsigned int max_segment_count; /* zero for unlimited */ unsigned long segment_boundary_mask; }; diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h index dfc099e..f380f79 100644 --- a/include/linux/dma-mapping.h +++ b/include/linux/dma-mapping.h @@ -111,6 +111,22 @@ static inline unsigned int dma_set_max_seg_size(struct device *dev, return -EIO; } +static inline unsigned int dma_get_max_seg_count(struct device *dev) +{ + return dev->dma_parms ? dev->dma_parms->max_segment_count : 0; +} + +static inline int dma_set_max_seg_count(struct device *dev, + unsigned int count) +{ + if (dev->dma_parms) { + dev->dma_parms->max_segment_count = count; + return 0; + } else + return -EIO; +} + + static inline unsigned long dma_get_seg_boundary(struct device *dev) { return dev->dma_parms ? -- 1.7.9.5

12 years, 10 months

2
1
0 0

Fwd: [Workshop-2011] Media summit at the Kernel Summit

by Hans Verkuil

FYI! If anyone is interested, then please discuss that on the linux-media mailinglist. I promised Mauro to forward it to a few mailinglists where this might be of interest, so here it is :-) Regards, Hans ---------- Forwarded Message ---------- Subject: [Workshop-2011] Media summit at the Kernel Summit - was: Fwd: Re: [Ksummit-2012-discuss] Organising Mini Summits within the Kernel Summit Date: Tue July 17 2012, 19:30:53 From: Mauro Carvalho Chehab <mchehab(a)redhat.com> To: Linux Media Mailing List <linux-media(a)vger.kernel.org>, workshop-2011(a)linuxtv.org As we did in 2012, we're planning to do a media summit again at KS/2012. The KS/2012 will happen in San Diego, CA, US, between Aug 26-28, just before the LinuxCon North America. In order to do it, I'd like to know who is interested on participate, and to get proposals about what subjects will be discussed there, in order to start planning the agenda. Thanks! Mauro

12 years, 10 months

1
0
0 0

[GIT PULL] CMA and DMA-mapping fixes for v3.5-rc7

by Marek Szyprowski

Hi Linus, I woukd like to ask you again for pulling another set of minor fixups for recently merged Contiguous Memory Allocator and ARM DMA-mapping changes. Those patches fix mysterious crashes on systems with CMA and Himem enabled as well as some corner cases caused by typical off-by-one bug. I'm sorry for a delay in sending this pull request which is caused by my vacations. The following changes since commit 6887a4131da3adaab011613776d865f4bcfb5678: Linux 3.5-rc5 (2012-06-30 16:08:57 -0700) with the top-most commit 46c87852e99cf8ce97e207b11cde19085837e39c ARM: dma-mapping: modify condition check while freeing pages are available in the git repository at: git://git.linaro.org/people/mszyprowski/linux-dma-mapping.git fixes-for-linus Marek Szyprowski (1): mm: cma: fix condition check when setting global cma area Prathyush K (1): ARM: dma-mapping: modify condition check while freeing pages Rabin Vincent (1): mm: cma: don't replace lowmem pages with highmem arch/arm/mm/dma-mapping.c | 4 ++-- include/asm-generic/dma-contiguous.h | 2 +- mm/page_alloc.c | 7 ++++++- 3 files changed, 9 insertions(+), 4 deletions(-) Thanks! Best regards Marek Szyprowski Samsung Poland R&D Center

12 years, 10 months

1
0
0 0

[RFC] dma-fence: dma-buf synchronization (v2)

by Rob Clark

From: Rob Clark <rob(a)ti.com> A dma-fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A dma-fence is transient, one-shot deal. It is allocated and attached to dma-buf's list of fences. When the one that attached it is done, with the pending operation, it can signal the fence removing it from the dma-buf's list of fences: + dma_buf_attach_fence() + dma_fence_signal() Other drivers can access the current fence on the dma-buf (if any), which increment's the fences refcnt: + dma_buf_get_fence() + dma_fence_put() The one pending on the fence can add an async callback (and optionally cancel it.. for example, to recover from GPU hangs): + dma_fence_add_callback() + dma_fence_cancel_callback() Or wait synchronously (optionally with timeout or from atomic context): + dma_fence_wait() A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = dma_buf_get_fence(dmabuf); if (fence->ops == &mem_dma_fence_ops) { dma_buf *fence_buf; mem_dma_fence_get_buf(fence, &fence_buf, &offset); ... tell the hw the memory location to wait on ... } else { /* fall-back to sw sync * / dma_fence_add_callback(fence, my_cb); } The memory location is itself backed by dma-buf, to simplify mapping to the device's address space, an idea borrowed from Maarten Lankhorst. NOTE: the memory location fence is not implemented yet, the above is just for explaining how it would work. On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. The other non-sw implementations would wrap the add/cancel_callback and wait fence ops, so that they can keep track if a device not supporting hw sync is waiting on the fence, and in this case should arrange to call dma_fence_signal() at some point after the condition has changed, to notify other devices waiting on the fence. If there are no sw waiters, this can be skipped to avoid waking the CPU unnecessarily. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw->hw signaling path (it can be handled same as sw->sw case), and therefore the fence->ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw->hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). --- drivers/base/Makefile | 2 +- drivers/base/dma-buf.c | 3 + drivers/base/dma-fence.c | 364 +++++++++++++++++++++++++++++++++++++++++++++ include/linux/dma-buf.h | 2 + include/linux/dma-fence.h | 128 ++++++++++++++++ 5 files changed, 498 insertions(+), 1 deletion(-) create mode 100644 drivers/base/dma-fence.c create mode 100644 include/linux/dma-fence.h diff --git a/drivers/base/Makefile b/drivers/base/Makefile index 5aa2d70..6e9f217 100644 --- a/drivers/base/Makefile +++ b/drivers/base/Makefile @@ -10,7 +10,7 @@ obj-$(CONFIG_CMA) += dma-contiguous.o obj-y += power/ obj-$(CONFIG_HAS_DMA) += dma-mapping.o obj-$(CONFIG_HAVE_GENERIC_DMA_COHERENT) += dma-coherent.o -obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o +obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o dma-fence.o obj-$(CONFIG_ISA) += isa.o obj-$(CONFIG_FW_LOADER) += firmware_class.o obj-$(CONFIG_NUMA) += node.o diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c index 24e88fe..b053236 100644 --- a/drivers/base/dma-buf.c +++ b/drivers/base/dma-buf.c @@ -39,6 +39,8 @@ static int dma_buf_release(struct inode *inode, struct file *file) dmabuf = file->private_data; + WARN_ON(!list_empty(&dmabuf->fence_list)); + dmabuf->ops->release(dmabuf); kfree(dmabuf); return 0; @@ -119,6 +121,7 @@ struct dma_buf *dma_buf_export(void *priv, const struct dma_buf_ops *ops, mutex_init(&dmabuf->lock); INIT_LIST_HEAD(&dmabuf->attachments); + INIT_LIST_HEAD(&dmabuf->fence_list); return dmabuf; } diff --git a/drivers/base/dma-fence.c b/drivers/base/dma-fence.c new file mode 100644 index 0000000..2bc25df --- /dev/null +++ b/drivers/base/dma-fence.c @@ -0,0 +1,364 @@ +/* + * Fence mechanism for dma-buf to allow for asynchronous dma access + * + * Copyright (C) 2012 Texas Instruments + * Author: Rob Clark <rob.clark(a)linaro.org> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <linux/slab.h> +#include <linux/sched.h> +#include <linux/export.h> +#include <linux/dma-fence.h> + +#define ATTACHED (1 << 0) +#define SIGNALED (1 << 1) + +/** + * dma_buf_attach_fence - Attach a fence to a dma-buf. + * + * @buf: the dma-buf to attach to + * @fence: the fence to attach + * + * A fence can only be attached to a single dma-buf. The dma-buf takes + * ownership of the fence, which is unref'd when the fence is signaled. + * The fence takes a reference to the dma-buf so the buffer will not be + * freed while there is a pending fence. + */ +int dma_buf_attach_fence(struct dma_buf *buf, struct dma_fence *fence) +{ + unsigned long flags; + int ret = -EINVAL; + + if (WARN_ON(!buf || !fence)) + return -EINVAL; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!fence->attached) { + get_dma_buf(buf); + fence->attached = true; + list_add(&fence->list_node, &buf->fence_list); + ret = 0; + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_buf_attach_fence); + +static void signal_fence(struct dma_fence *fence) +{ + list_del(&fence->list_node); + wake_up_all_locked(&fence->event_queue); +} + +/** + * dma_fence_signal - Signal a fence. + * + * @fence: The fence to signal + * + * All registered callbacks will be called directly (synchronously) and + * all blocked waters will be awoken. + * + * TODO: any value in adding a dma_fence_cancel(), for example to recov + * from hung gpu? It would behave like dma_fence_signal() but return + * an error to waiters and cb's to let them know that the condition they + * are waiting for will never happen. + */ +int dma_fence_signal(struct dma_fence *fence) +{ + unsigned long flags; + int ret = -EINVAL; + + if (WARN_ON(!fence)) + return -EINVAL; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (fence->attached && !fence->signaled) { + fence->signaled = true; + signal_fence(fence); + ret = 0; + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + dma_fence_put(fence); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_signal); + +/** + * dma_buf_get_fence - Get the most recent pending fence attached to the + * dma-buf. + * + * @buf: the dma-buf whose fence to get + * + * If this returns NULL, there are no pending fences. Otherwise this + * takes a reference to the returned fence, so the caller must later + * call dma_fence_put() to release the reference. + */ +struct dma_fence *dma_buf_get_fence(struct dma_buf *buf) +{ + struct dma_fence *fence = NULL; + unsigned long flags; + + if (WARN_ON(!buf)) + return ERR_PTR(-EINVAL); + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!list_empty(&buf->fence_list)) { + fence = list_first_entry(&buf->fence_list, + struct dma_fence, list_node); + dma_fence_get(fence); + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return fence; +} +EXPORT_SYMBOL_GPL(dma_buf_get_fence); + +static void release_fence(struct kref *kref) +{ + struct dma_fence *fence = + container_of(kref, struct dma_fence, refcount); + + WARN_ON(waitqueue_active(&fence->event_queue)); + + kfree(fence); +} + +/** + * dma_fence_put - Release a reference to the fence. + */ +void dma_fence_put(struct dma_fence *fence) +{ + WARN_ON(!fence); + kref_put(&fence->refcount, release_fence); +} +EXPORT_SYMBOL_GPL(dma_fence_put); + +/** + * dma_fence_get - Take a reference to the fence. + * + * In most cases this is used only internally by dma-fence. + */ +void dma_fence_get(struct dma_fence *fence) +{ + WARN_ON(!fence); + kref_get(&fence->refcount); +} +EXPORT_SYMBOL_GPL(dma_fence_get); + +static int check_signaling(struct dma_fence *fence) +{ + bool enable_signaling = false; + unsigned long flags; + int ret = 0; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (!fence->needs_sw_signal && !fence->signaled) + enable_signaling = fence->needs_sw_signal = true; + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + if (enable_signaling) + ret = fence->ops->enable_signaling(fence); + + return ret; +} + +/** + * dma_fence_add_callback - Add a callback to be called when the fence + * is signaled. + * + * @fence: The fence to wait on + * @cb: The callback to register + * + * Any number of callbacks can be registered to a fence, but a callback + * can only be registered to once fence at a time. + * + * Note that the callback can be called from an atomic context. If + * fence is already signaled, this function will return -ENOENT (and + * *not* call the callback) + */ +int dma_fence_add_callback(struct dma_fence *fence, + struct dma_fence_cb *cb) +{ + unsigned long flags; + int ret; + + if (WARN_ON(!fence || !cb)) + return -EINVAL; + + ret = check_signaling(fence); + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (ret == -ENOENT) { + /* if state changed while we dropped the lock, dispatch now */ + signal_fence(fence); + } else if (!fence->signaled && !ret) { + dma_fence_get(fence); + cb->fence = fence; + __add_wait_queue(&fence->event_queue, &cb->base); + ret = 0; + } else { + ret = -EINVAL; + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_add_callback); + +/** + * dma_fence_cancel_callback - Remove a previously registered callback. + * + * @cb: The callback to unregister + * + * The callback will not be called after this function returns, but could + * be called before this function returns. + */ +int dma_fence_cancel_callback(struct dma_fence_cb *cb) +{ + struct dma_fence *fence; + unsigned long flags; + int ret = -EINVAL; + + if (WARN_ON(!cb)) + return -EINVAL; + + fence = cb->fence; + + spin_lock_irqsave(&fence->event_queue.lock, flags); + if (fence) { + __remove_wait_queue(&fence->event_queue, &cb->base); + cb->fence = NULL; + dma_fence_put(fence); + ret = 0; + } + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_cancel_callback); + +/** + * dma_fence_wait - Wait for a fence to be signaled. + * + * @fence: The fence to wait on + * @interruptible: if true, do an interruptible wait + * @timeout: timeout, in jiffies + * + * Returns -ENOENT if the fence has already passed. + */ +int dma_fence_wait(struct dma_fence *fence, bool interruptible, long timeout) +{ + unsigned long flags; + int ret; + + if (WARN_ON(!fence)) + return -EINVAL; + + ret = check_signaling(fence); + if (ret == -ENOENT) { + spin_lock_irqsave(&fence->event_queue.lock, flags); + signal_fence(fence); + spin_unlock_irqrestore(&fence->event_queue.lock, flags); + return ret; + } + + if (ret) + return ret; + + if (interruptible) + ret = wait_event_interruptible_timeout(fence->event_queue, + fence->signaled, timeout); + else + ret = wait_event_timeout(fence->event_queue, + fence->signaled, timeout); + + return ret; +} +EXPORT_SYMBOL_GPL(dma_fence_wait); + +int __dma_fence_wake_func(wait_queue_t *wait, unsigned mode, + int flags, void *key) +{ + struct dma_fence_cb *cb = + container_of(wait, struct dma_fence_cb, base); + struct dma_fence *fence = cb->fence; + int ret; + + ret = cb->func(cb, fence); + cb->fence = NULL; + dma_fence_put(fence); + + return ret; +} +EXPORT_SYMBOL_GPL(__dma_fence_wake_func); + +/* + * Helpers intended to be used by the ops of the dma_fence implementation: + * + * NOTE: helpers and fxns intended to be used by other dma-fence + * implementations are not exported.. I'm not really sure if it makes + * sense to have a dma-fence implementation that is itself a module. + */ + +void __dma_fence_init(struct dma_fence *fence, struct dma_fence_ops *ops) +{ + WARN_ON(!ops || !ops->enable_signaling); + + kref_init(&fence->refcount); + fence->ops = ops; + init_waitqueue_head(&fence->event_queue); +} + +/* + * Pure sw implementation for dma-fence. The CPU always gets involved. + */ + +static int sw_enable_signaling(struct dma_fence *fence) +{ + /* + * pure sw, no irq's to enable, because the fence creator will + * always call dma_fence_signal() + */ + return 0; +} + +static struct dma_fence_ops sw_fence_ops = { + .enable_signaling = sw_enable_signaling, +}; + +/** + * dma_fence_create - Create a simple sw-only fence. + * + * This fence only supports signaling from/to CPU. Other implementations + * of dma-fence can be used to support hardware to hardware signaling, if + * supported by the hardware, and use the dma_fence_helper_* functions for + * compatibility with other devices that only support sw signaling. + */ +struct dma_fence *dma_fence_create(void) +{ + struct dma_fence *fence; + + fence = kzalloc(sizeof(struct dma_fence), GFP_KERNEL); + if (!fence) + return ERR_PTR(-ENOMEM); + + __dma_fence_init(fence, &sw_fence_ops); + + return fence; +} +EXPORT_SYMBOL_GPL(dma_fence_create); diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h index eb48f38..ff4ec6c 100644 --- a/include/linux/dma-buf.h +++ b/include/linux/dma-buf.h @@ -122,6 +122,8 @@ struct dma_buf { /* mutex to serialize list manipulation and attach/detach */ struct mutex lock; void *priv; + /* list of pending dma_fence's */ + struct list_head fence_list; }; /** diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h new file mode 100644 index 0000000..336afd2 --- /dev/null +++ b/include/linux/dma-fence.h @@ -0,0 +1,128 @@ +/* + * Fence mechanism for dma-buf to allow for asynchronous dma access + * + * Copyright (C) 2012 Texas Instruments + * Author: Rob Clark <rob.clark(a)linaro.org> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#ifndef __DMA_FENCE_H__ +#define __DMA_FENCE_H__ + +#include <linux/err.h> +#include <linux/list.h> +#include <linux/wait.h> +#include <linux/list.h> +#include <linux/dma-buf.h> + +struct dma_fence; +struct dma_fence_ops; +struct dma_fence_cb; + +struct dma_fence { + struct kref refcount; + struct dma_fence_ops *ops; + wait_queue_head_t event_queue; + struct list_head list_node; /* for dmabuf's fence_list */ + + /* has this fence been attached to a dma-buf yet? */ + bool attached : 1; + + /* has this fence been signaled yet? */ + bool signaled : 1; + + /* do we have one or more waiters or callbacks? */ + bool needs_sw_signal : 1; +}; + +typedef int (*dma_fence_func_t)(struct dma_fence_cb *cb, + struct dma_fence *fence); + +struct dma_fence_cb { + wait_queue_t base; + dma_fence_func_t func; + + /* + * This is initialized when the cb is added, and NULL'd when it + * is canceled or expired, so can be used to for error checking + * if the cb is already pending. A dma_fence_cb can be pending + * on at most one fence at a time. + */ + struct dma_fence *fence; +}; + +struct dma_fence_ops { + /** + * For fence implementations that have the capability for hw->hw + * signaling, they can implement this op to enable the necessary + * irqs, or insert commands into cmdstream, etc. This is called + * in the first wait() or add_callback() path to let the fence + * implementation know that there is another driver waiting on + * the signal (ie. hw->sw case). + * + * A return value of -ENOENT will indicate that the fence has + * already passed. + */ + int (*enable_signaling)(struct dma_fence *fence); +}; + +int __dma_fence_wake_func(wait_queue_t *wait, unsigned mode, + int flags, void *key); + +#define DMA_FENCE_CB_INITIALIZER(cb_func) { \ + .base = { .func = __dma_fence_wake_func }, \ + .func = (cb_func), \ + } + +#define DECLARE_DMA_FENCE(name, cb_func) \ + struct dma_fence_cb name = DMA_FENCE_CB_INITIALIZER(cb_func) + + +/* + * TODO does it make sense to be able to enable dma-fence without dma-buf, + * or visa versa? + */ +#ifdef CONFIG_DMA_SHARED_BUFFER + +int dma_buf_attach_fence(struct dma_buf *buf, struct dma_fence *fence); +struct dma_fence *dma_buf_get_fence(struct dma_buf *buf); + +/* create a basic (pure sw) fence: */ +struct dma_fence *dma_fence_create(void); + +/* intended to be used by other dma_fence implementations: */ +void __dma_fence_init(struct dma_fence *fence, struct dma_fence_ops *ops); + +void dma_fence_get(struct dma_fence *fence); +void dma_fence_put(struct dma_fence *fence); +int dma_fence_signal(struct dma_fence *fence); + +int dma_fence_add_callback(struct dma_fence *fence, + struct dma_fence_cb *cb); +int dma_fence_cancel_callback(struct dma_fence_cb *cb); +int dma_fence_wait(struct dma_fence *fence, bool interruptible, long timeout); + +/* helpers intended to be used by the ops of the dma_fence implementation: */ +int dma_fence_helper_signal(struct dma_fence *fence); +int dma_fence_helper_add_callback(struct dma_fence *fence, + struct dma_fence_cb *cb); +int dma_fence_helper_cancel_callback(struct dma_fence_cb *cb); +int dma_fence_helper_wait(struct dma_fence *fence, bool interruptible, + long timeout); + +#else +// TODO +#endif /* CONFIG_DMA_SHARED_BUFFER */ + +#endif /* __DMA_FENCE_H__ */ -- 1.7.9.5

12 years, 10 months

3
2
0 0

[PATCH v2] ARM: dma-mapping: modify condition check while freeing pages

by Prathyush K

WARNING: at mm/vmalloc.c:1471 __iommu_free_buffer+0xcc/0xd0() Trying to vfree() nonexistent vm area (ef095000) Modules linked in: [<c0015a18>] (unwind_backtrace+0x0/0xfc) from [<c0025a94>] (warn_slowpath_common+0x54/0x64) [<c0025a94>] (warn_slowpath_common+0x54/0x64) from [<c0025b38>] (warn_slowpath_fmt+0x30/0x40) [<c0025b38>] (warn_slowpath_fmt+0x30/0x40) from [<c0016de0>] (__iommu_free_buffer+0xcc/0xd0) [<c0016de0>] (__iommu_free_buffer+0xcc/0xd0) from [<c0229a5c>] (exynos_drm_free_buf+0xe4/0x138) [<c0229a5c>] (exynos_drm_free_buf+0xe4/0x138) from [<c022b358>] (exynos_drm_gem_destroy+0x80/0xfc) [<c022b358>] (exynos_drm_gem_destroy+0x80/0xfc) from [<c0211230>] (drm_gem_object_free+0x28/0x34) [<c0211230>] (drm_gem_object_free+0x28/0x34) from [<c0211bd0>] (drm_gem_object_release_handle+0xcc/0xd8) [<c0211bd0>] (drm_gem_object_release_handle+0xcc/0xd8) from [<c01abe10>] (idr_for_each+0x74/0xb8) [<c01abe10>] (idr_for_each+0x74/0xb8) from [<c02114e4>] (drm_gem_release+0x1c/0x30) [<c02114e4>] (drm_gem_release+0x1c/0x30) from [<c0210ae8>] (drm_release+0x608/0x694) [<c0210ae8>] (drm_release+0x608/0x694) from [<c00b75a0>] (fput+0xb8/0x228) [<c00b75a0>] (fput+0xb8/0x228) from [<c00b40c4>] (filp_close+0x64/0x84) [<c00b40c4>] (filp_close+0x64/0x84) from [<c0029d54>] (put_files_struct+0xe8/0x104) [<c0029d54>] (put_files_struct+0xe8/0x104) from [<c002b930>] (do_exit+0x608/0x774) [<c002b930>] (do_exit+0x608/0x774) from [<c002bae4>] (do_group_exit+0x48/0xb4) [<c002bae4>] (do_group_exit+0x48/0xb4) from [<c002bb60>] (sys_exit_group+0x10/0x18) [<c002bb60>] (sys_exit_group+0x10/0x18) from [<c000ee80>] (ret_fast_syscall+0x0/0x30) This patch modifies the condition while freeing to match the condition used while allocation. This fixes the above warning which arises when array size is equal to PAGE_SIZE where allocation is done using kzalloc but free is done using vfree. Signed-off-by: Prathyush K <prathyush.k(a)samsung.com> --- arch/arm/mm/dma-mapping.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c index dc560dc..0f9358b 100644 --- a/arch/arm/mm/dma-mapping.c +++ b/arch/arm/mm/dma-mapping.c @@ -1091,7 +1091,7 @@ error: while (--i) if (pages[i]) __free_pages(pages[i], 0); - if (array_size < PAGE_SIZE) + if (array_size <= PAGE_SIZE) kfree(pages); else vfree(pages); @@ -1106,7 +1106,7 @@ static int __iommu_free_buffer(struct device *dev, struct page **pages, size_t s for (i = 0; i < count; i++) if (pages[i]) __free_pages(pages[i], 0); - if (array_size < PAGE_SIZE) + if (array_size <= PAGE_SIZE) kfree(pages); else vfree(pages); -- 1.7.0.4

12 years, 10 months

2
1
0 0

Re: [Linaro-mm-sig] [RFC] dma-fence: dma-buf synchronization (v2)

by Rob Clark

On Fri, Jul 13, 2012 at 12:35 PM, Tom Cooksey <tom.cooksey(a)arm.com> wrote: > My other thought is around atomicity. Could this be extended to > (safely) allow for hardware devices which might want to access > multiple buffers simultaneously? I think it probably can with > some tweaks to the interface? An atomic function which does > something like "give me all the fences for all these buffers > and add this fence to each instead/as-well-as"? fwiw, what I'm leaning towards right now is combining dma-fence w/ Maarten's idea of dma-buf-mgr (not sure if you saw his patches?). And let dmabufmgr handle the multi-buffer reservation stuff. And possibly the read vs write access, although this I'm not 100% sure on... the other option being the concept of read vs write (or exclusive/non-exclusive) fences. In the current state, the fence is quite simple, and doesn't care *what* it is fencing, which seems advantageous when you get into trying to deal with combinations of devices sharing buffers, some of whom can do hw sync, and some who can't. So having a bit of partitioning from the code dealing w/ sequencing who can access the buffers when and for what purpose seems like it might not be a bad idea. Although I'm still working through the different alternatives. BR, -R

12 years, 10 months

2
2
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

Linaro-mm-sig