From 5237e95f63761477b7ea45499d08b89383a77eab Mon Sep 17 00:00:00 2001 From: Robin Murphy Date: Mon, 24 Jul 2017 18:29:27 +0100 Subject: [PATCH 01/12] dma-mapping: reduce dma_mapping_error inline bloat Thanks to the nested inlining, all drivers correctly calling dma_mapping_error() after a mapping a page or single buffer generate two calls to get_arch_dma_ops() per callsite, which all adds up to a fair old chunk of useless code, e.g. ~3KB for an arm64 defconfig plus extras: text data bss dec hex filename 13051391 1503898 327768 14883057 e318f1 vmlinux.o.old 13050751 1503898 327768 14882417 e31671 vmlinux.o.new Give the compiler a hand by making it clear we want the same ops. Signed-off-by: Robin Murphy Reviewed-by: Marek Szyprowski Signed-off-by: Christoph Hellwig --- include/linux/dma-mapping.h | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h index 03c0196a6f24..66d8ea68f40b 100644 --- a/include/linux/dma-mapping.h +++ b/include/linux/dma-mapping.h @@ -565,10 +565,11 @@ static inline void dma_free_noncoherent(struct device *dev, size_t size, static inline int dma_mapping_error(struct device *dev, dma_addr_t dma_addr) { - debug_dma_mapping_error(dev, dma_addr); + const struct dma_map_ops *ops = get_dma_ops(dev); - if (get_dma_ops(dev)->mapping_error) - return get_dma_ops(dev)->mapping_error(dev, dma_addr); + debug_dma_mapping_error(dev, dma_addr); + if (ops->mapping_error) + return ops->mapping_error(dev, dma_addr); return 0; } From d64bfe9f70e265ed7dd5453e0a31a9ecc9ca475f Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 14 Jun 2017 10:27:43 +0200 Subject: [PATCH 02/12] sgiseeq: switch to dma_alloc_attrs Use dma_alloc_attrs directly instead of the dma_alloc_noncoherent wrapper. Signed-off-by: Christoph Hellwig Acked-by: Ralf Baechle Acked-by: David S. Miller --- drivers/net/ethernet/seeq/sgiseeq.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/net/ethernet/seeq/sgiseeq.c b/drivers/net/ethernet/seeq/sgiseeq.c index 70347720fdf9..573691bc3b71 100644 --- a/drivers/net/ethernet/seeq/sgiseeq.c +++ b/drivers/net/ethernet/seeq/sgiseeq.c @@ -737,8 +737,8 @@ static int sgiseeq_probe(struct platform_device *pdev) sp = netdev_priv(dev); /* Make private data page aligned */ - sr = dma_alloc_noncoherent(&pdev->dev, sizeof(*sp->srings), - &sp->srings_dma, GFP_KERNEL); + sr = dma_alloc_attrs(&pdev->dev, sizeof(*sp->srings), &sp->srings_dma, + GFP_KERNEL, DMA_ATTR_NON_CONSISTENT); if (!sr) { printk(KERN_ERR "Sgiseeq: Page alloc failed, aborting.\n"); err = -ENOMEM; @@ -813,8 +813,8 @@ static int sgiseeq_remove(struct platform_device *pdev) struct sgiseeq_private *sp = netdev_priv(dev); unregister_netdev(dev); - dma_free_noncoherent(&pdev->dev, sizeof(*sp->srings), sp->srings, - sp->srings_dma); + dma_free_attrs(&pdev->dev, sizeof(*sp->srings), sp->srings, + sp->srings_dma, DMA_ATTR_NON_CONSISTENT); free_netdev(dev); return 0; From ae9f608f07e7828801430a00f67882fc386b33d5 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 14 Jun 2017 10:29:47 +0200 Subject: [PATCH 03/12] au1000_eth: switch to dma_alloc_attrs Use dma_alloc_attrs directly instead of the dma_alloc_noncoherent wrapper. Signed-off-by: Christoph Hellwig Acked-by: David S. Miller --- drivers/net/ethernet/amd/au1000_eth.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) diff --git a/drivers/net/ethernet/amd/au1000_eth.c b/drivers/net/ethernet/amd/au1000_eth.c index a3c90fe5de00..73ca8879ada7 100644 --- a/drivers/net/ethernet/amd/au1000_eth.c +++ b/drivers/net/ethernet/amd/au1000_eth.c @@ -1180,9 +1180,10 @@ static int au1000_probe(struct platform_device *pdev) /* Allocate the data buffers * Snooping works fine with eth on all au1xxx */ - aup->vaddr = (u32)dma_alloc_noncoherent(NULL, MAX_BUF_SIZE * - (NUM_TX_BUFFS + NUM_RX_BUFFS), - &aup->dma_addr, 0); + aup->vaddr = (u32)dma_alloc_attrs(NULL, MAX_BUF_SIZE * + (NUM_TX_BUFFS + NUM_RX_BUFFS), + &aup->dma_addr, 0, + DMA_ATTR_NON_CONSISTENT); if (!aup->vaddr) { dev_err(&pdev->dev, "failed to allocate data buffers\n"); err = -ENOMEM; @@ -1361,8 +1362,9 @@ err_remap3: err_remap2: iounmap(aup->mac); err_remap1: - dma_free_noncoherent(NULL, MAX_BUF_SIZE * (NUM_TX_BUFFS + NUM_RX_BUFFS), - (void *)aup->vaddr, aup->dma_addr); + dma_free_attrs(NULL, MAX_BUF_SIZE * (NUM_TX_BUFFS + NUM_RX_BUFFS), + (void *)aup->vaddr, aup->dma_addr, + DMA_ATTR_NON_CONSISTENT); err_vaddr: free_netdev(dev); err_alloc: @@ -1394,9 +1396,9 @@ static int au1000_remove(struct platform_device *pdev) if (aup->tx_db_inuse[i]) au1000_ReleaseDB(aup, aup->tx_db_inuse[i]); - dma_free_noncoherent(NULL, MAX_BUF_SIZE * - (NUM_TX_BUFFS + NUM_RX_BUFFS), - (void *)aup->vaddr, aup->dma_addr); + dma_free_attrs(NULL, MAX_BUF_SIZE * (NUM_TX_BUFFS + NUM_RX_BUFFS), + (void *)aup->vaddr, aup->dma_addr, + DMA_ATTR_NON_CONSISTENT); iounmap(aup->macdma); iounmap(aup->mac); From 7f683b920479530150cafbed9dc02eeafb90486b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 14 Jun 2017 10:33:48 +0200 Subject: [PATCH 04/12] i825xx: switch to switch to dma_alloc_attrs This way we can always pass DMA_ATTR_NON_CONSISTENT, the SNI mips version will simply ignore the flag. Signed-off-by: Christoph Hellwig Acked-by: David S. Miller --- drivers/net/ethernet/i825xx/lasi_82596.c | 6 ++---- drivers/net/ethernet/i825xx/lib82596.c | 9 +++++---- drivers/net/ethernet/i825xx/sni_82596.c | 6 ++---- 3 files changed, 9 insertions(+), 12 deletions(-) diff --git a/drivers/net/ethernet/i825xx/lasi_82596.c b/drivers/net/ethernet/i825xx/lasi_82596.c index d787fdd5db7b..d5b5021aa759 100644 --- a/drivers/net/ethernet/i825xx/lasi_82596.c +++ b/drivers/net/ethernet/i825xx/lasi_82596.c @@ -96,8 +96,6 @@ #define OPT_SWAP_PORT 0x0001 /* Need to wordswp on the MPU port */ -#define DMA_ALLOC dma_alloc_noncoherent -#define DMA_FREE dma_free_noncoherent #define DMA_WBACK(ndev, addr, len) \ do { dma_cache_sync((ndev)->dev.parent, (void *)addr, len, DMA_TO_DEVICE); } while (0) @@ -200,8 +198,8 @@ static int lan_remove_chip(struct parisc_device *pdev) struct i596_private *lp = netdev_priv(dev); unregister_netdev (dev); - DMA_FREE(&pdev->dev, sizeof(struct i596_private), - (void *)lp->dma, lp->dma_addr); + dma_free_attrs(&pdev->dev, sizeof(struct i596_private), lp->dma, + lp->dma_addr, DMA_ATTR_NON_CONSISTENT); free_netdev (dev); return 0; } diff --git a/drivers/net/ethernet/i825xx/lib82596.c b/drivers/net/ethernet/i825xx/lib82596.c index 8449c58f01fd..f00a1dc2128c 100644 --- a/drivers/net/ethernet/i825xx/lib82596.c +++ b/drivers/net/ethernet/i825xx/lib82596.c @@ -1063,8 +1063,9 @@ static int i82596_probe(struct net_device *dev) if (!dev->base_addr || !dev->irq) return -ENODEV; - dma = (struct i596_dma *) DMA_ALLOC(dev->dev.parent, - sizeof(struct i596_dma), &lp->dma_addr, GFP_KERNEL); + dma = dma_alloc_attrs(dev->dev.parent, sizeof(struct i596_dma), + &lp->dma_addr, GFP_KERNEL, + DMA_ATTR_NON_CONSISTENT); if (!dma) { printk(KERN_ERR "%s: Couldn't get shared memory\n", __FILE__); return -ENOMEM; @@ -1085,8 +1086,8 @@ static int i82596_probe(struct net_device *dev) i = register_netdev(dev); if (i) { - DMA_FREE(dev->dev.parent, sizeof(struct i596_dma), - (void *)dma, lp->dma_addr); + dma_free_attrs(dev->dev.parent, sizeof(struct i596_dma), + dma, lp->dma_addr, DMA_ATTR_NON_CONSISTENT); return i; } diff --git a/drivers/net/ethernet/i825xx/sni_82596.c b/drivers/net/ethernet/i825xx/sni_82596.c index 2af7f77345fb..b2c04a789744 100644 --- a/drivers/net/ethernet/i825xx/sni_82596.c +++ b/drivers/net/ethernet/i825xx/sni_82596.c @@ -23,8 +23,6 @@ static const char sni_82596_string[] = "snirm_82596"; -#define DMA_ALLOC dma_alloc_coherent -#define DMA_FREE dma_free_coherent #define DMA_WBACK(priv, addr, len) do { } while (0) #define DMA_INV(priv, addr, len) do { } while (0) #define DMA_WBACK_INV(priv, addr, len) do { } while (0) @@ -152,8 +150,8 @@ static int sni_82596_driver_remove(struct platform_device *pdev) struct i596_private *lp = netdev_priv(dev); unregister_netdev(dev); - DMA_FREE(dev->dev.parent, sizeof(struct i596_private), - lp->dma, lp->dma_addr); + dma_free_attrs(dev->dev.parent, sizeof(struct i596_private), lp->dma, + lp->dma_addr, DMA_ATTR_NON_CONSISTENT); iounmap(lp->ca); iounmap(lp->mpu_port); free_netdev (dev); From 2fd523c57e520899e05ec663d04743005bcef5b2 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 25 Aug 2017 17:06:13 +0200 Subject: [PATCH 05/12] dma-mapping: remove dma_alloc_noncoherent and dma_free_noncoherent No users left, everyone switched to the _attrs versions. Signed-off-by: Christoph Hellwig --- Documentation/DMA-API.txt | 30 +++++++++++++++------------- arch/metag/include/asm/dma-mapping.h | 2 +- arch/nios2/include/asm/dma-mapping.h | 2 +- arch/tile/include/asm/dma-mapping.h | 4 ++-- include/linux/dma-mapping.h | 14 ------------- 5 files changed, 20 insertions(+), 32 deletions(-) diff --git a/Documentation/DMA-API.txt b/Documentation/DMA-API.txt index 45b29326d719..ef3a04fcad65 100644 --- a/Documentation/DMA-API.txt +++ b/Documentation/DMA-API.txt @@ -515,14 +515,15 @@ API at all. :: void * - dma_alloc_noncoherent(struct device *dev, size_t size, - dma_addr_t *dma_handle, gfp_t flag) + dma_alloc_attrs(struct device *dev, size_t size, dma_addr_t *dma_handle, + gfp_t flag, unsigned long attrs) -Identical to dma_alloc_coherent() except that the platform will -choose to return either consistent or non-consistent memory as it sees -fit. By using this API, you are guaranteeing to the platform that you -have all the correct and necessary sync points for this memory in the -driver should it choose to return non-consistent memory. +Identical to dma_alloc_coherent() except that when the +DMA_ATTR_NON_CONSISTENT flags is passed in the attrs argument, the +platform will choose to return either consistent or non-consistent memory +as it sees fit. By using this API, you are guaranteeing to the platform +that you have all the correct and necessary sync points for this memory +in the driver should it choose to return non-consistent memory. Note: where the platform can return consistent memory, it will guarantee that the sync points become nops. @@ -535,12 +536,13 @@ that simply cannot make consistent memory. :: void - dma_free_noncoherent(struct device *dev, size_t size, void *cpu_addr, - dma_addr_t dma_handle) + dma_free_attrs(struct device *dev, size_t size, void *cpu_addr, + dma_addr_t dma_handle, unsigned long attrs) -Free memory allocated by the nonconsistent API. All parameters must -be identical to those passed in (and returned by -dma_alloc_noncoherent()). +Free memory allocated by the dma_alloc_attrs(). All parameters common +parameters must identical to those otherwise passed to dma_fre_coherent, +and the attrs argument must be identical to the attrs passed to +dma_alloc_attrs(). :: @@ -564,8 +566,8 @@ memory or doing partial flushes. dma_cache_sync(struct device *dev, void *vaddr, size_t size, enum dma_data_direction direction) -Do a partial sync of memory that was allocated by -dma_alloc_noncoherent(), starting at virtual address vaddr and +Do a partial sync of memory that was allocated by dma_alloc_attrs() with +the DMA_ATTR_NON_CONSISTENT flag starting at virtual address vaddr and continuing on for size. Again, you *must* observe the cache line boundaries when doing this. diff --git a/arch/metag/include/asm/dma-mapping.h b/arch/metag/include/asm/dma-mapping.h index fad3dc3cb210..ea573be2b6d0 100644 --- a/arch/metag/include/asm/dma-mapping.h +++ b/arch/metag/include/asm/dma-mapping.h @@ -9,7 +9,7 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus) } /* - * dma_alloc_noncoherent() returns non-cacheable memory, so there's no need to + * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to * do any flushing here. */ static inline void diff --git a/arch/nios2/include/asm/dma-mapping.h b/arch/nios2/include/asm/dma-mapping.h index 7b3c6f280293..f8dc62222741 100644 --- a/arch/nios2/include/asm/dma-mapping.h +++ b/arch/nios2/include/asm/dma-mapping.h @@ -18,7 +18,7 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus) } /* - * dma_alloc_noncoherent() returns non-cacheable memory, so there's no need to + * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to * do any flushing here. */ static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size, diff --git a/arch/tile/include/asm/dma-mapping.h b/arch/tile/include/asm/dma-mapping.h index bbc71a29b2c6..7061dc8af43a 100644 --- a/arch/tile/include/asm/dma-mapping.h +++ b/arch/tile/include/asm/dma-mapping.h @@ -68,8 +68,8 @@ static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size) int dma_set_mask(struct device *dev, u64 mask); /* - * dma_alloc_noncoherent() is #defined to return coherent memory, - * so there's no need to do any flushing here. + * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to + * do any flushing here. */ static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size, enum dma_data_direction direction) diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h index 66d8ea68f40b..4c98cc96971f 100644 --- a/include/linux/dma-mapping.h +++ b/include/linux/dma-mapping.h @@ -549,20 +549,6 @@ static inline void dma_free_coherent(struct device *dev, size_t size, return dma_free_attrs(dev, size, cpu_addr, dma_handle, 0); } -static inline void *dma_alloc_noncoherent(struct device *dev, size_t size, - dma_addr_t *dma_handle, gfp_t gfp) -{ - return dma_alloc_attrs(dev, size, dma_handle, gfp, - DMA_ATTR_NON_CONSISTENT); -} - -static inline void dma_free_noncoherent(struct device *dev, size_t size, - void *cpu_addr, dma_addr_t dma_handle) -{ - dma_free_attrs(dev, size, cpu_addr, dma_handle, - DMA_ATTR_NON_CONSISTENT); -} - static inline int dma_mapping_error(struct device *dev, dma_addr_t dma_addr) { const struct dma_map_ops *ops = get_dma_ops(dev); From 723288836628bc1c0855f3bb7b64b1803e4b9e4a Mon Sep 17 00:00:00 2001 From: Robin Murphy Date: Thu, 31 Aug 2017 11:32:54 +0100 Subject: [PATCH 06/12] of: restrict DMA configuration Moving DMA configuration to happen later at driver probe time had the unnoticed side-effect that we now perform DMA configuration for *every* device represented in DT, rather than only those explicitly created by the of_platform and PCI code. As Christoph points out, this is not really the best thing to do. Whilst there may well be other DMA-capable buses that can benefit from having their children automatically configured after the bridge has probed, there are also plenty of others like USB, MDIO, etc. that definitely do not support DMA and should not be indiscriminately processed. The good news is that in most cases the DT "dma-ranges" property serves as an appropriate indicator - per a strict interpretation of the spec, anything lacking a "dma-ranges" property should be considered not to have a mapping of DMA address space from its children to its parent, thus anything for which of_dma_get_range() does not succeed does not need DMA configuration. Certain bus types have a general expectation of DMA capability and carry a well-established precedent that an absent "dma-ranges" implies the same as the empty property, so we automatically opt those in to DMA configuration regardless, to avoid regressing most existing platforms. Fixes: 09515ef5ddad ("of/acpi: Configure dma operations at probe time for platform/amba/pci bus devices") Reported-by: Christoph Hellwig Signed-off-by: Robin Murphy Acked-by: Rob Herring Signed-off-by: Christoph Hellwig --- drivers/of/device.c | 48 ++++++++++++++++++++++++++++++--------------- 1 file changed, 32 insertions(+), 16 deletions(-) diff --git a/drivers/of/device.c b/drivers/of/device.c index e0a28ea341fe..04c4c952dc57 100644 --- a/drivers/of/device.c +++ b/drivers/of/device.c @@ -9,6 +9,9 @@ #include #include #include +#include +#include +#include #include #include "of_private.h" @@ -84,31 +87,28 @@ int of_device_add(struct platform_device *ofdev) */ int of_dma_configure(struct device *dev, struct device_node *np) { - u64 dma_addr, paddr, size; + u64 dma_addr, paddr, size = 0; int ret; bool coherent; unsigned long offset; const struct iommu_ops *iommu; u64 mask; - /* - * Set default coherent_dma_mask to 32 bit. Drivers are expected to - * setup the correct supported mask. - */ - if (!dev->coherent_dma_mask) - dev->coherent_dma_mask = DMA_BIT_MASK(32); - - /* - * Set it to coherent_dma_mask by default if the architecture - * code has not set it. - */ - if (!dev->dma_mask) - dev->dma_mask = &dev->coherent_dma_mask; - ret = of_dma_get_range(np, &dma_addr, &paddr, &size); if (ret < 0) { + /* + * For legacy reasons, we have to assume some devices need + * DMA configuration regardless of whether "dma-ranges" is + * correctly specified or not. + */ + if (!dev_is_pci(dev) && +#ifdef CONFIG_ARM_AMBA + dev->bus != &amba_bustype && +#endif + dev->bus != &platform_bus_type) + return ret == -ENODEV ? 0 : ret; + dma_addr = offset = 0; - size = max(dev->coherent_dma_mask, dev->coherent_dma_mask + 1); } else { offset = PFN_DOWN(paddr - dma_addr); @@ -129,6 +129,22 @@ int of_dma_configure(struct device *dev, struct device_node *np) dev_dbg(dev, "dma_pfn_offset(%#08lx)\n", offset); } + /* + * Set default coherent_dma_mask to 32 bit. Drivers are expected to + * setup the correct supported mask. + */ + if (!dev->coherent_dma_mask) + dev->coherent_dma_mask = DMA_BIT_MASK(32); + /* + * Set it to coherent_dma_mask by default if the architecture + * code has not set it. + */ + if (!dev->dma_mask) + dev->dma_mask = &dev->coherent_dma_mask; + + if (!size) + size = max(dev->coherent_dma_mask, dev->coherent_dma_mask + 1); + dev->dma_pfn_offset = offset; /* From b32dbc1e0bf0eaa8bdd725491b361d5fc2f57a85 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 25 Aug 2017 17:12:18 +0200 Subject: [PATCH 07/12] dma-coherent: remove the DMA_MEMORY_INCLUDES_CHILDREN flag This flag was never implemented or used. Signed-off-by: Christoph Hellwig Reviewed-by: Robin Murphy --- Documentation/DMA-API.txt | 4 ---- drivers/base/dma-coherent.c | 2 -- drivers/char/virtio_console.c | 3 --- include/linux/dma-mapping.h | 3 +-- 4 files changed, 1 insertion(+), 11 deletions(-) diff --git a/Documentation/DMA-API.txt b/Documentation/DMA-API.txt index ef3a04fcad65..dddf52e768d6 100644 --- a/Documentation/DMA-API.txt +++ b/Documentation/DMA-API.txt @@ -600,10 +600,6 @@ flags can be ORed together and are: One or both of these flags must be present. -- DMA_MEMORY_INCLUDES_CHILDREN - make the declared memory be allocated by - dma_alloc_coherent of any child devices of this one (for memory residing - on a bridge). - - DMA_MEMORY_EXCLUSIVE - only allocate memory from the declared regions. Do not allow dma_alloc_coherent() to fall back to system memory when it's out of memory in the declared region. diff --git a/drivers/base/dma-coherent.c b/drivers/base/dma-coherent.c index 1c152aed6b82..6f6dc4d41788 100644 --- a/drivers/base/dma-coherent.c +++ b/drivers/base/dma-coherent.c @@ -109,8 +109,6 @@ static int dma_assign_coherent_memory(struct device *dev, return -EBUSY; dev->dma_mem = mem; - /* FIXME: this routine just ignores DMA_MEMORY_INCLUDES_CHILDREN */ - return 0; } diff --git a/drivers/char/virtio_console.c b/drivers/char/virtio_console.c index ad843eb02ae7..48e86ce9fac8 100644 --- a/drivers/char/virtio_console.c +++ b/drivers/char/virtio_console.c @@ -451,9 +451,6 @@ static struct port_buffer *alloc_buf(struct virtqueue *vq, size_t buf_size, * device is created by remoteproc, the DMA memory is * associated with the grandparent device: * vdev => rproc => platform-dev. - * The code here would have been less quirky if - * DMA_MEMORY_INCLUDES_CHILDREN had been supported - * in dma-coherent.c */ if (!vq->vdev->dev.parent || !vq->vdev->dev.parent->parent) goto free_buf; diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h index 4c98cc96971f..abf369b11f54 100644 --- a/include/linux/dma-mapping.h +++ b/include/linux/dma-mapping.h @@ -696,8 +696,7 @@ static inline int dma_get_cache_alignment(void) /* flags for the coherent memory api */ #define DMA_MEMORY_MAP 0x01 #define DMA_MEMORY_IO 0x02 -#define DMA_MEMORY_INCLUDES_CHILDREN 0x04 -#define DMA_MEMORY_EXCLUSIVE 0x08 +#define DMA_MEMORY_EXCLUSIVE 0x04 #ifdef CONFIG_HAVE_GENERIC_DMA_COHERENT int dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr, From 2436bdcda53ff4abb7897c87fa29ef3de8055344 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 25 Aug 2017 17:13:09 +0200 Subject: [PATCH 08/12] dma-coherent: remove the DMA_MEMORY_MAP and DMA_MEMORY_IO flags DMA_MEMORY_IO was never used in the tree, so remove it. That means there is no need for the DMA_MEMORY_MAP flag either now, so remove it as well and change dma_declare_coherent_memory to return a normal errno value. Signed-off-by: Christoph Hellwig Reviewed-by: Marek Szyprowski --- Documentation/DMA-API.txt | 21 +-------- arch/arm/mach-imx/mach-imx27_visstrim_m10.c | 44 +++++++----------- arch/arm/mach-imx/mach-mx31moboard.c | 12 ++--- arch/sh/drivers/pci/fixups-dreamcast.c | 3 +- drivers/base/dma-coherent.c | 46 ++++++------------- drivers/base/dma-mapping.c | 7 +-- .../soc_camera/sh_mobile_ceu_camera.c | 5 +- drivers/scsi/NCR_Q720.c | 3 +- drivers/usb/host/ohci-sm501.c | 7 ++- drivers/usb/host/ohci-tmio.c | 9 ++-- include/linux/dma-mapping.h | 6 +-- 11 files changed, 52 insertions(+), 111 deletions(-) diff --git a/Documentation/DMA-API.txt b/Documentation/DMA-API.txt index dddf52e768d6..ac66ae2509a9 100644 --- a/Documentation/DMA-API.txt +++ b/Documentation/DMA-API.txt @@ -592,30 +592,11 @@ size is the size of the area (must be multiples of PAGE_SIZE). flags can be ORed together and are: -- DMA_MEMORY_MAP - request that the memory returned from - dma_alloc_coherent() be directly writable. - -- DMA_MEMORY_IO - request that the memory returned from - dma_alloc_coherent() be addressable using read()/write()/memcpy_toio() etc. - -One or both of these flags must be present. - - DMA_MEMORY_EXCLUSIVE - only allocate memory from the declared regions. Do not allow dma_alloc_coherent() to fall back to system memory when it's out of memory in the declared region. -The return value will be either DMA_MEMORY_MAP or DMA_MEMORY_IO and -must correspond to a passed in flag (i.e. no returning DMA_MEMORY_IO -if only DMA_MEMORY_MAP were passed in) for success or zero for -failure. - -Note, for DMA_MEMORY_IO returns, all subsequent memory returned by -dma_alloc_coherent() may no longer be accessed directly, but instead -must be accessed using the correct bus functions. If your driver -isn't prepared to handle this contingency, it should not specify -DMA_MEMORY_IO in the input flags. - -As a simplification for the platforms, only **one** such region of +As a simplification for the platforms, only *one* such region of memory may be declared per device. For reasons of efficiency, most platforms choose to track the declared diff --git a/arch/arm/mach-imx/mach-imx27_visstrim_m10.c b/arch/arm/mach-imx/mach-imx27_visstrim_m10.c index dd75a4756761..5169dfba9718 100644 --- a/arch/arm/mach-imx/mach-imx27_visstrim_m10.c +++ b/arch/arm/mach-imx/mach-imx27_visstrim_m10.c @@ -245,7 +245,6 @@ static phys_addr_t mx2_camera_base __initdata; static void __init visstrim_analog_camera_init(void) { struct platform_device *pdev; - int dma; gpio_set_value(TVP5150_PWDN, 1); ndelay(1); @@ -258,12 +257,9 @@ static void __init visstrim_analog_camera_init(void) if (IS_ERR(pdev)) return; - dma = dma_declare_coherent_memory(&pdev->dev, - mx2_camera_base, mx2_camera_base, - MX2_CAMERA_BUF_SIZE, - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE); - if (!(dma & DMA_MEMORY_MAP)) - return; + dma_declare_coherent_memory(&pdev->dev, mx2_camera_base, + mx2_camera_base, MX2_CAMERA_BUF_SIZE, + DMA_MEMORY_EXCLUSIVE); } static void __init visstrim_reserve(void) @@ -444,16 +440,13 @@ static const struct imx_ssi_platform_data visstrim_m10_ssi_pdata __initconst = { static void __init visstrim_coda_init(void) { struct platform_device *pdev; - int dma; pdev = imx27_add_coda(); - dma = dma_declare_coherent_memory(&pdev->dev, - mx2_camera_base + MX2_CAMERA_BUF_SIZE, - mx2_camera_base + MX2_CAMERA_BUF_SIZE, - MX2_CAMERA_BUF_SIZE, - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE); - if (!(dma & DMA_MEMORY_MAP)) - return; + dma_declare_coherent_memory(&pdev->dev, + mx2_camera_base + MX2_CAMERA_BUF_SIZE, + mx2_camera_base + MX2_CAMERA_BUF_SIZE, + MX2_CAMERA_BUF_SIZE, + DMA_MEMORY_EXCLUSIVE); } /* DMA deinterlace */ @@ -466,24 +459,21 @@ static void __init visstrim_deinterlace_init(void) { int ret = -ENOMEM; struct platform_device *pdev = &visstrim_deinterlace; - int dma; ret = platform_device_register(pdev); - dma = dma_declare_coherent_memory(&pdev->dev, - mx2_camera_base + 2 * MX2_CAMERA_BUF_SIZE, - mx2_camera_base + 2 * MX2_CAMERA_BUF_SIZE, - MX2_CAMERA_BUF_SIZE, - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE); - if (!(dma & DMA_MEMORY_MAP)) - return; + dma_declare_coherent_memory(&pdev->dev, + mx2_camera_base + 2 * MX2_CAMERA_BUF_SIZE, + mx2_camera_base + 2 * MX2_CAMERA_BUF_SIZE, + MX2_CAMERA_BUF_SIZE, + DMA_MEMORY_EXCLUSIVE); } /* Emma-PrP for format conversion */ static void __init visstrim_emmaprp_init(void) { struct platform_device *pdev; - int dma; + int ret; pdev = imx27_add_mx2_emmaprp(); if (IS_ERR(pdev)) @@ -493,11 +483,11 @@ static void __init visstrim_emmaprp_init(void) * Use the same memory area as the analog camera since both * devices are, by nature, exclusive. */ - dma = dma_declare_coherent_memory(&pdev->dev, + ret = dma_declare_coherent_memory(&pdev->dev, mx2_camera_base, mx2_camera_base, MX2_CAMERA_BUF_SIZE, - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE); - if (!(dma & DMA_MEMORY_MAP)) + DMA_MEMORY_EXCLUSIVE); + if (ret) pr_err("Failed to declare memory for emmaprp\n"); } diff --git a/arch/arm/mach-imx/mach-mx31moboard.c b/arch/arm/mach-imx/mach-mx31moboard.c index bde9a9af6714..3cd030960c98 100644 --- a/arch/arm/mach-imx/mach-mx31moboard.c +++ b/arch/arm/mach-imx/mach-mx31moboard.c @@ -475,7 +475,7 @@ static phys_addr_t mx3_camera_base __initdata; static int __init mx31moboard_init_cam(void) { - int dma, ret = -ENOMEM; + int dma, ret; struct platform_device *pdev; imx31_add_ipu_core(); @@ -484,11 +484,11 @@ static int __init mx31moboard_init_cam(void) if (IS_ERR(pdev)) return PTR_ERR(pdev); - dma = dma_declare_coherent_memory(&pdev->dev, - mx3_camera_base, mx3_camera_base, - MX3_CAMERA_BUF_SIZE, - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE); - if (!(dma & DMA_MEMORY_MAP)) + ret = dma_declare_coherent_memory(&pdev->dev, + mx3_camera_base, mx3_camera_base, + MX3_CAMERA_BUF_SIZE, + DMA_MEMORY_EXCLUSIVE); + if (ret) goto err; ret = platform_device_add(pdev); diff --git a/arch/sh/drivers/pci/fixups-dreamcast.c b/arch/sh/drivers/pci/fixups-dreamcast.c index 1d1c5a227e50..c931d5872ffe 100644 --- a/arch/sh/drivers/pci/fixups-dreamcast.c +++ b/arch/sh/drivers/pci/fixups-dreamcast.c @@ -63,11 +63,10 @@ static void gapspci_fixup_resources(struct pci_dev *dev) res.end = GAPSPCI_DMA_BASE + GAPSPCI_DMA_SIZE - 1; res.flags = IORESOURCE_MEM; pcibios_resource_to_bus(dev->bus, ®ion, &res); - BUG_ON(!dma_declare_coherent_memory(&dev->dev, + BUG_ON(dma_declare_coherent_memory(&dev->dev, res.start, region.start, resource_size(&res), - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE)); break; default: diff --git a/drivers/base/dma-coherent.c b/drivers/base/dma-coherent.c index 6f6dc4d41788..b275eeb739f1 100644 --- a/drivers/base/dma-coherent.c +++ b/drivers/base/dma-coherent.c @@ -46,15 +46,10 @@ static bool dma_init_coherent_memory( int pages = size >> PAGE_SHIFT; int bitmap_size = BITS_TO_LONGS(pages) * sizeof(long); - if ((flags & (DMA_MEMORY_MAP | DMA_MEMORY_IO)) == 0) - goto out; if (!size) goto out; - if (flags & DMA_MEMORY_MAP) - mem_base = memremap(phys_addr, size, MEMREMAP_WC); - else - mem_base = ioremap(phys_addr, size); + mem_base = memremap(phys_addr, size, MEMREMAP_WC); if (!mem_base) goto out; @@ -77,12 +72,8 @@ static bool dma_init_coherent_memory( out: kfree(dma_mem); - if (mem_base) { - if (flags & DMA_MEMORY_MAP) - memunmap(mem_base); - else - iounmap(mem_base); - } + if (mem_base) + memunmap(mem_base); return false; } @@ -91,10 +82,7 @@ static void dma_release_coherent_memory(struct dma_coherent_mem *mem) if (!mem) return; - if (mem->flags & DMA_MEMORY_MAP) - memunmap(mem->virt_base); - else - iounmap(mem->virt_base); + memunmap(mem->virt_base); kfree(mem->bitmap); kfree(mem); } @@ -116,16 +104,16 @@ int dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr, dma_addr_t device_addr, size_t size, int flags) { struct dma_coherent_mem *mem; + int ret; - if (!dma_init_coherent_memory(phys_addr, device_addr, size, flags, - &mem)) - return 0; + ret = dma_init_coherent_memory(phys_addr, device_addr, size, flags, &mem); + if (ret) + return ret; - if (dma_assign_coherent_memory(dev, mem) == 0) - return flags & DMA_MEMORY_MAP ? DMA_MEMORY_MAP : DMA_MEMORY_IO; - - dma_release_coherent_memory(mem); - return 0; + ret = dma_assign_coherent_memory(dev, mem); + if (ret) + dma_release_coherent_memory(mem); + return ret; } EXPORT_SYMBOL(dma_declare_coherent_memory); @@ -186,15 +174,9 @@ static void *__dma_alloc_from_coherent(struct dma_coherent_mem *mem, */ *dma_handle = mem->device_base + (pageno << PAGE_SHIFT); ret = mem->virt_base + (pageno << PAGE_SHIFT); - dma_memory_map = (mem->flags & DMA_MEMORY_MAP); spin_unlock_irqrestore(&mem->spinlock, flags); - if (dma_memory_map) - memset(ret, 0, size); - else - memset_io(ret, 0, size); - + memset(ret, 0, size); return ret; - err: spin_unlock_irqrestore(&mem->spinlock, flags); return NULL; @@ -360,7 +342,7 @@ static int rmem_dma_device_init(struct reserved_mem *rmem, struct device *dev) if (!mem && !dma_init_coherent_memory(rmem->base, rmem->base, rmem->size, - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE, + DMA_MEMORY_EXCLUSIVE, &mem)) { pr_err("Reserved memory: failed to init DMA memory pool at %pa, size %ld MiB\n", &rmem->base, (unsigned long)rmem->size / SZ_1M); diff --git a/drivers/base/dma-mapping.c b/drivers/base/dma-mapping.c index b555ff9dd8fc..e584eddef0a7 100644 --- a/drivers/base/dma-mapping.c +++ b/drivers/base/dma-mapping.c @@ -176,13 +176,10 @@ int dmam_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr, rc = dma_declare_coherent_memory(dev, phys_addr, device_addr, size, flags); - if (rc) { + if (!rc) devres_add(dev, res); - rc = 0; - } else { + else devres_free(res); - rc = -ENOMEM; - } return rc; } diff --git a/drivers/media/platform/soc_camera/sh_mobile_ceu_camera.c b/drivers/media/platform/soc_camera/sh_mobile_ceu_camera.c index 96dc01750bc0..36762ec954e7 100644 --- a/drivers/media/platform/soc_camera/sh_mobile_ceu_camera.c +++ b/drivers/media/platform/soc_camera/sh_mobile_ceu_camera.c @@ -1708,11 +1708,10 @@ static int sh_mobile_ceu_probe(struct platform_device *pdev) err = dma_declare_coherent_memory(&pdev->dev, res->start, res->start, resource_size(res), - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE); - if (!err) { + if (err) { dev_err(&pdev->dev, "Unable to declare CEU memory.\n"); - return -ENXIO; + return err; } pcdev->video_limit = resource_size(res); diff --git a/drivers/scsi/NCR_Q720.c b/drivers/scsi/NCR_Q720.c index 05835bf1bf9c..54e7d26908ee 100644 --- a/drivers/scsi/NCR_Q720.c +++ b/drivers/scsi/NCR_Q720.c @@ -217,8 +217,7 @@ NCR_Q720_probe(struct device *dev) } if (dma_declare_coherent_memory(dev, base_addr, base_addr, - mem_size, DMA_MEMORY_MAP) - != DMA_MEMORY_MAP) { + mem_size, 0)) { printk(KERN_ERR "NCR_Q720: DMA declare memory failed\n"); goto out_release_region; } diff --git a/drivers/usb/host/ohci-sm501.c b/drivers/usb/host/ohci-sm501.c index a8b8d8b8d9f3..d4e0f7cd96fa 100644 --- a/drivers/usb/host/ohci-sm501.c +++ b/drivers/usb/host/ohci-sm501.c @@ -123,13 +123,12 @@ static int ohci_hcd_sm501_drv_probe(struct platform_device *pdev) * regular memory. The HCD_LOCAL_MEM flag does just that. */ - if (!dma_declare_coherent_memory(dev, mem->start, + retval = dma_declare_coherent_memory(dev, mem->start, mem->start - mem->parent->start, resource_size(mem), - DMA_MEMORY_MAP | - DMA_MEMORY_EXCLUSIVE)) { + DMA_MEMORY_EXCLUSIVE); + if (retval) { dev_err(dev, "cannot declare coherent memory\n"); - retval = -ENXIO; goto err1; } diff --git a/drivers/usb/host/ohci-tmio.c b/drivers/usb/host/ohci-tmio.c index cfcfadfc94fc..16d081a093bb 100644 --- a/drivers/usb/host/ohci-tmio.c +++ b/drivers/usb/host/ohci-tmio.c @@ -227,13 +227,10 @@ static int ohci_hcd_tmio_drv_probe(struct platform_device *dev) goto err_ioremap_regs; } - if (!dma_declare_coherent_memory(&dev->dev, sram->start, - sram->start, - resource_size(sram), - DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE)) { - ret = -EBUSY; + ret = dma_declare_coherent_memory(&dev->dev, sram->start, sram->start, + resource_size(sram), DMA_MEMORY_EXCLUSIVE); + if (ret) goto err_dma_declare; - } if (cell->enable) { ret = cell->enable(dev); diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h index abf369b11f54..b7dd11c02a45 100644 --- a/include/linux/dma-mapping.h +++ b/include/linux/dma-mapping.h @@ -694,9 +694,7 @@ static inline int dma_get_cache_alignment(void) #endif /* flags for the coherent memory api */ -#define DMA_MEMORY_MAP 0x01 -#define DMA_MEMORY_IO 0x02 -#define DMA_MEMORY_EXCLUSIVE 0x04 +#define DMA_MEMORY_EXCLUSIVE 0x01 #ifdef CONFIG_HAVE_GENERIC_DMA_COHERENT int dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr, @@ -709,7 +707,7 @@ static inline int dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr, dma_addr_t device_addr, size_t size, int flags) { - return 0; + return -ENOSYS; } static inline void From 4a75682ce3bcd57d146db25b888e71f659ba4998 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 26 Aug 2017 11:25:02 +0200 Subject: [PATCH 09/12] MAINTAINERS: use the iommu list for the dma-mapping subsystem Maintaining a subsystem with linux-kernel as the main list is painful as it has way to much traffic. On the other hand the dma-mapping subsystem is small enough that a list on its own would be silly. So use the list for the closes subsystem instead instead. Signed-off-by: Christoph Hellwig Acked-by: Robin Murphy Acked-by: Joerg Roedel Acked-by: Marek Szyprowski --- MAINTAINERS | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/MAINTAINERS b/MAINTAINERS index 6f7721d1634c..1df11ed346a7 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -4175,7 +4175,7 @@ DMA MAPPING HELPERS M: Christoph Hellwig M: Marek Szyprowski R: Robin Murphy -L: linux-kernel@vger.kernel.org +L: iommu@lists.linux-foundation.org T: git git://git.infradead.org/users/hch/dma-mapping.git W: http://git.infradead.org/users/hch/dma-mapping.git S: Supported From 34bbf1335fac345eb69a13d143d2dcad368c7b56 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 4 Sep 2017 08:29:48 +0200 Subject: [PATCH 10/12] dma-coherent: remove an unused variable Signed-off-by: Christoph Hellwig Reported-by: Stephen Rothwell --- drivers/base/dma-coherent.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/base/dma-coherent.c b/drivers/base/dma-coherent.c index b275eeb739f1..f82a504583d4 100644 --- a/drivers/base/dma-coherent.c +++ b/drivers/base/dma-coherent.c @@ -157,7 +157,6 @@ static void *__dma_alloc_from_coherent(struct dma_coherent_mem *mem, int order = get_order(size); unsigned long flags; int pageno; - int dma_memory_map; void *ret; spin_lock_irqsave(&mem->spinlock, flags); From edeb8e4ccb79eaed4a5126945cce00489f09b849 Mon Sep 17 00:00:00 2001 From: Fabio Estevam Date: Mon, 4 Sep 2017 11:58:14 -0300 Subject: [PATCH 11/12] ARM: imx: mx31moboard: Remove unused 'dma' variable Commit 2436bdcda53f ("dma-coherent: remove the DMA_MEMORY_MAP and DMA_MEMORY_IO flags") missed to remove the 'dma' variable causing the following build warning: arch/arm/mach-imx/mach-mx31moboard.c:478:6: warning: unused variable 'dma' [-Wunused-variable] Remove the unused 'dma' variable. Signed-off-by: Fabio Estevam Signed-off-by: Christoph Hellwig --- arch/arm/mach-imx/mach-mx31moboard.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/arm/mach-imx/mach-mx31moboard.c b/arch/arm/mach-imx/mach-mx31moboard.c index 3cd030960c98..7716f83aecdd 100644 --- a/arch/arm/mach-imx/mach-mx31moboard.c +++ b/arch/arm/mach-imx/mach-mx31moboard.c @@ -475,7 +475,7 @@ static phys_addr_t mx3_camera_base __initdata; static int __init mx31moboard_init_cam(void) { - int dma, ret; + int ret; struct platform_device *pdev; imx31_add_ipu_core(); From d35b0996fef3bfe76665e87bbff7d95c6807350a Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Tue, 5 Sep 2017 10:10:09 +0200 Subject: [PATCH 12/12] dma-coherent: fix dma_declare_coherent_memory() logic error A recent change interprets the return code of dma_init_coherent_memory as an error value, but it is instead a boolean, where 'true' indicates success. This leads causes the caller to always do the wrong thing, and also triggers a compile-time warning about it: drivers/base/dma-coherent.c: In function 'dma_declare_coherent_memory': drivers/base/dma-coherent.c:99:15: error: 'mem' may be used uninitialized in this function [-Werror=maybe-uninitialized] I ended up changing the code a little more, to give use the usual error handling, as this seemed the best way to fix up the warning and make the code look reasonable at the same time. Fixes: 2436bdcda53f ("dma-coherent: remove the DMA_MEMORY_MAP and DMA_MEMORY_IO flags") Signed-off-by: Arnd Bergmann Signed-off-by: Christoph Hellwig --- drivers/base/dma-coherent.c | 38 ++++++++++++++++++++++++------------- 1 file changed, 25 insertions(+), 13 deletions(-) diff --git a/drivers/base/dma-coherent.c b/drivers/base/dma-coherent.c index f82a504583d4..a39b2166b145 100644 --- a/drivers/base/dma-coherent.c +++ b/drivers/base/dma-coherent.c @@ -37,7 +37,7 @@ static inline dma_addr_t dma_get_device_base(struct device *dev, return mem->device_base; } -static bool dma_init_coherent_memory( +static int dma_init_coherent_memory( phys_addr_t phys_addr, dma_addr_t device_addr, size_t size, int flags, struct dma_coherent_mem **mem) { @@ -45,20 +45,28 @@ static bool dma_init_coherent_memory( void __iomem *mem_base = NULL; int pages = size >> PAGE_SHIFT; int bitmap_size = BITS_TO_LONGS(pages) * sizeof(long); + int ret; - if (!size) + if (!size) { + ret = -EINVAL; goto out; + } mem_base = memremap(phys_addr, size, MEMREMAP_WC); - if (!mem_base) + if (!mem_base) { + ret = -EINVAL; goto out; - + } dma_mem = kzalloc(sizeof(struct dma_coherent_mem), GFP_KERNEL); - if (!dma_mem) + if (!dma_mem) { + ret = -ENOMEM; goto out; + } dma_mem->bitmap = kzalloc(bitmap_size, GFP_KERNEL); - if (!dma_mem->bitmap) + if (!dma_mem->bitmap) { + ret = -ENOMEM; goto out; + } dma_mem->virt_base = mem_base; dma_mem->device_base = device_addr; @@ -68,13 +76,13 @@ static bool dma_init_coherent_memory( spin_lock_init(&dma_mem->spinlock); *mem = dma_mem; - return true; + return 0; out: kfree(dma_mem); if (mem_base) memunmap(mem_base); - return false; + return ret; } static void dma_release_coherent_memory(struct dma_coherent_mem *mem) @@ -338,14 +346,18 @@ static struct reserved_mem *dma_reserved_default_memory __initdata; static int rmem_dma_device_init(struct reserved_mem *rmem, struct device *dev) { struct dma_coherent_mem *mem = rmem->priv; + int ret; - if (!mem && - !dma_init_coherent_memory(rmem->base, rmem->base, rmem->size, - DMA_MEMORY_EXCLUSIVE, - &mem)) { + if (!mem) + return -ENODEV; + + ret = dma_init_coherent_memory(rmem->base, rmem->base, rmem->size, + DMA_MEMORY_EXCLUSIVE, &mem); + + if (ret) { pr_err("Reserved memory: failed to init DMA memory pool at %pa, size %ld MiB\n", &rmem->base, (unsigned long)rmem->size / SZ_1M); - return -ENODEV; + return ret; } mem->use_dev_dma_pfn_offset = true; rmem->priv = mem;