/* Copyright (c) 2002,2007-2015, The Linux Foundation. All rights reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 and * only 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. * */ #include #include #include #include #include #include #include #include #include #include "kgsl.h" #include "kgsl_sharedmem.h" #include "kgsl_cffdump.h" #include "kgsl_device.h" #include "kgsl_log.h" /* * The user can set this from debugfs to force failed memory allocations to * fail without trying OOM first. This is a debug setting useful for * stress applications that want to test failure cases without pushing the * system into unrecoverable OOM panics */ static bool sharedmem_noretry_flag; static DEFINE_MUTEX(kernel_map_global_lock); struct cp2_mem_chunks { unsigned int chunk_list; unsigned int chunk_list_size; unsigned int chunk_size; } __attribute__ ((__packed__)); struct cp2_lock_req { struct cp2_mem_chunks chunks; unsigned int mem_usage; unsigned int lock; } __attribute__ ((__packed__)); #define MEM_PROTECT_LOCK_ID2 0x0A #define MEM_PROTECT_LOCK_ID2_FLAT 0x11 /* An attribute for showing per-process memory statistics */ struct kgsl_mem_entry_attribute { struct attribute attr; int memtype; ssize_t (*show)(struct kgsl_process_private *priv, int type, char *buf); }; #define to_mem_entry_attr(a) \ container_of(a, struct kgsl_mem_entry_attribute, attr) #define __MEM_ENTRY_ATTR(_type, _name, _show) \ { \ .attr = { .name = __stringify(_name), .mode = 0444 }, \ .memtype = _type, \ .show = _show, \ } /* * A structure to hold the attributes for a particular memory type. * For each memory type in each process we store the current and maximum * memory usage and display the counts in sysfs. This structure and * the following macro allow us to simplify the definition for those * adding new memory types */ struct mem_entry_stats { int memtype; struct kgsl_mem_entry_attribute attr; struct kgsl_mem_entry_attribute max_attr; }; #define MEM_ENTRY_STAT(_type, _name) \ { \ .memtype = _type, \ .attr = __MEM_ENTRY_ATTR(_type, _name, mem_entry_show), \ .max_attr = __MEM_ENTRY_ATTR(_type, _name##_max, \ mem_entry_max_show), \ } static void kgsl_cma_unlock_secure(struct kgsl_memdesc *memdesc); /** * Show the current amount of memory allocated for the given memtype */ static ssize_t mem_entry_show(struct kgsl_process_private *priv, int type, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", priv->stats[type].cur); } /** * Show the maximum memory allocated for the given memtype through the life of * the process */ static ssize_t mem_entry_max_show(struct kgsl_process_private *priv, int type, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", priv->stats[type].max); } static ssize_t mem_entry_sysfs_show(struct kobject *kobj, struct attribute *attr, char *buf) { struct kgsl_mem_entry_attribute *pattr = to_mem_entry_attr(attr); struct kgsl_process_private *priv; ssize_t ret; /* * 1. sysfs_remove_file waits for reads to complete before the node * is deleted. * 2. kgsl_process_init_sysfs takes a refcount to the process_private, * which is put at the end of kgsl_process_uninit_sysfs. * These two conditions imply that priv will not be freed until this * function completes, and no further locking is needed. */ priv = kobj ? container_of(kobj, struct kgsl_process_private, kobj) : NULL; if (priv && pattr->show) ret = pattr->show(priv, pattr->memtype, buf); else ret = -EIO; return ret; } static const struct sysfs_ops mem_entry_sysfs_ops = { .show = mem_entry_sysfs_show, }; static struct kobj_type ktype_mem_entry = { .sysfs_ops = &mem_entry_sysfs_ops, }; static struct mem_entry_stats mem_stats[] = { MEM_ENTRY_STAT(KGSL_MEM_ENTRY_KERNEL, kernel), MEM_ENTRY_STAT(KGSL_MEM_ENTRY_USER, user), #ifdef CONFIG_ION MEM_ENTRY_STAT(KGSL_MEM_ENTRY_ION, ion), #endif }; void kgsl_process_uninit_sysfs(struct kgsl_process_private *private) { int i; for (i = 0; i < ARRAY_SIZE(mem_stats); i++) { sysfs_remove_file(&private->kobj, &mem_stats[i].attr.attr); sysfs_remove_file(&private->kobj, &mem_stats[i].max_attr.attr); } kobject_put(&private->kobj); /* Put the refcount we got in kgsl_process_init_sysfs */ kgsl_process_private_put(private); } /** * kgsl_process_init_sysfs() - Initialize and create sysfs files for a process * * @device: Pointer to kgsl device struct * @private: Pointer to the structure for the process * * kgsl_process_init_sysfs() is called at the time of creating the * process struct when a process opens the kgsl device for the first time. * This function creates the sysfs files for the process. */ void kgsl_process_init_sysfs(struct kgsl_device *device, struct kgsl_process_private *private) { unsigned char name[16]; int i; /* Keep private valid until the sysfs enries are removed. */ kgsl_process_private_get(private); snprintf(name, sizeof(name), "%d", private->pid); if (kobject_init_and_add(&private->kobj, &ktype_mem_entry, kgsl_driver.prockobj, name)) { WARN(1, "Unable to add sysfs dir '%s'\n", name); return; } for (i = 0; i < ARRAY_SIZE(mem_stats); i++) { if (sysfs_create_file(&private->kobj, &mem_stats[i].attr.attr)) WARN(1, "Couldn't create sysfs file '%s'\n", mem_stats[i].attr.attr.name); if (sysfs_create_file(&private->kobj, &mem_stats[i].max_attr.attr)) WARN(1, "Couldn't create sysfs file '%s'\n", mem_stats[i].max_attr.attr.name); } } static ssize_t kgsl_drv_memstat_show(struct device *dev, struct device_attribute *attr, char *buf) { uint64_t val = 0; if (!strcmp(attr->attr.name, "vmalloc")) val = atomic_long_read(&kgsl_driver.stats.vmalloc); else if (!strcmp(attr->attr.name, "vmalloc_max")) val = atomic_long_read(&kgsl_driver.stats.vmalloc_max); else if (!strcmp(attr->attr.name, "page_alloc")) val = atomic_long_read(&kgsl_driver.stats.page_alloc); else if (!strcmp(attr->attr.name, "page_alloc_max")) val = atomic_long_read(&kgsl_driver.stats.page_alloc_max); else if (!strcmp(attr->attr.name, "coherent")) val = atomic_long_read(&kgsl_driver.stats.coherent); else if (!strcmp(attr->attr.name, "coherent_max")) val = atomic_long_read(&kgsl_driver.stats.coherent_max); else if (!strcmp(attr->attr.name, "secure")) val = atomic_long_read(&kgsl_driver.stats.secure); else if (!strcmp(attr->attr.name, "secure_max")) val = atomic_long_read(&kgsl_driver.stats.secure_max); else if (!strcmp(attr->attr.name, "mapped")) val = atomic_long_read(&kgsl_driver.stats.mapped); else if (!strcmp(attr->attr.name, "mapped_max")) val = atomic_long_read(&kgsl_driver.stats.mapped_max); return snprintf(buf, PAGE_SIZE, "%llu\n", val); } static ssize_t kgsl_drv_full_cache_threshold_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { int ret; unsigned int thresh = 0; ret = kgsl_sysfs_store(buf, &thresh); if (ret) return ret; kgsl_driver.full_cache_threshold = thresh; return count; } static ssize_t kgsl_drv_full_cache_threshold_show(struct device *dev, struct device_attribute *attr, char *buf) { return snprintf(buf, PAGE_SIZE, "%d\n", kgsl_driver.full_cache_threshold); } static DEVICE_ATTR(vmalloc, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(vmalloc_max, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(page_alloc, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(page_alloc_max, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(coherent, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(coherent_max, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(secure, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(secure_max, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(mapped, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(mapped_max, 0444, kgsl_drv_memstat_show, NULL); static DEVICE_ATTR(full_cache_threshold, 0644, kgsl_drv_full_cache_threshold_show, kgsl_drv_full_cache_threshold_store); static const struct device_attribute *drv_attr_list[] = { &dev_attr_vmalloc, &dev_attr_vmalloc_max, &dev_attr_page_alloc, &dev_attr_page_alloc_max, &dev_attr_coherent, &dev_attr_coherent_max, &dev_attr_secure, &dev_attr_secure_max, &dev_attr_mapped, &dev_attr_mapped_max, &dev_attr_full_cache_threshold, NULL }; void kgsl_sharedmem_uninit_sysfs(void) { kgsl_remove_device_sysfs_files(&kgsl_driver.virtdev, drv_attr_list); } int kgsl_sharedmem_init_sysfs(void) { return kgsl_create_device_sysfs_files(&kgsl_driver.virtdev, drv_attr_list); } static int kgsl_allocate_secure(struct kgsl_device *device, struct kgsl_memdesc *memdesc, struct kgsl_pagetable *pagetable, uint64_t size) { int ret; if (MMU_FEATURE(&device->mmu, KGSL_MMU_HYP_SECURE_ALLOC)) ret = kgsl_sharedmem_page_alloc_user(memdesc, pagetable, size); else ret = kgsl_cma_alloc_secure(device, memdesc, size); return ret; } int kgsl_allocate_user(struct kgsl_device *device, struct kgsl_memdesc *memdesc, struct kgsl_pagetable *pagetable, uint64_t size, uint64_t mmapsize, uint64_t flags) { int ret; if (size == 0) return -EINVAL; memdesc->flags = flags; if (kgsl_mmu_get_mmutype() == KGSL_MMU_TYPE_NONE) ret = kgsl_cma_alloc_coherent(device, memdesc, pagetable, size); else if (flags & KGSL_MEMFLAGS_SECURE) ret = kgsl_allocate_secure(device, memdesc, pagetable, size); else ret = kgsl_sharedmem_page_alloc_user(memdesc, pagetable, size); return ret; } static int kgsl_page_alloc_vmfault(struct kgsl_memdesc *memdesc, struct vm_area_struct *vma, struct vm_fault *vmf) { int i, pgoff; struct scatterlist *s = memdesc->sgt->sgl; unsigned int offset; offset = ((unsigned long) vmf->virtual_address - vma->vm_start); if (offset >= memdesc->size) return VM_FAULT_SIGBUS; pgoff = offset >> PAGE_SHIFT; /* * The sglist might be comprised of mixed blocks of memory depending * on how many 64K pages were allocated. This means we have to do math * to find the actual 4K page to map in user space */ for (i = 0; i < memdesc->sgt->nents; i++) { int npages = s->length >> PAGE_SHIFT; if (pgoff < npages) { struct page *page = sg_page(s); page = nth_page(page, pgoff); get_page(page); vmf->page = page; return 0; } pgoff -= npages; s = sg_next(s); } return VM_FAULT_SIGBUS; } /* * kgsl_page_alloc_unmap_kernel() - Unmap the memory in memdesc * * @memdesc: The memory descriptor which contains information about the memory * * Unmaps the memory mapped into kernel address space */ static void kgsl_page_alloc_unmap_kernel(struct kgsl_memdesc *memdesc) { mutex_lock(&kernel_map_global_lock); if (!memdesc->hostptr) { BUG_ON(memdesc->hostptr_count); goto done; } memdesc->hostptr_count--; if (memdesc->hostptr_count) goto done; vunmap(memdesc->hostptr); atomic_long_sub(memdesc->size, &kgsl_driver.stats.vmalloc); memdesc->hostptr = NULL; done: mutex_unlock(&kernel_map_global_lock); } static void kgsl_page_alloc_free(struct kgsl_memdesc *memdesc) { unsigned int i = 0; struct scatterlist *sg; kgsl_page_alloc_unmap_kernel(memdesc); /* we certainly do not expect the hostptr to still be mapped */ BUG_ON(memdesc->hostptr); /* Secure buffers need to be unlocked before being freed */ if (memdesc->priv & KGSL_MEMDESC_TZ_LOCKED) { int ret; int dest_perms = PERM_READ | PERM_WRITE | PERM_EXEC; int source_vm = VMID_CP_PIXEL; int dest_vm = VMID_HLOS; ret = hyp_assign_table(memdesc->sgt, &source_vm, 1, &dest_vm, &dest_perms, 1); if (ret) { pr_err("Secure buf unlock failed: gpuaddr: %llx size: %llx ret: %d\n", memdesc->gpuaddr, memdesc->size, ret); BUG(); } atomic_long_sub(memdesc->size, &kgsl_driver.stats.secure); } else { atomic_long_sub(memdesc->size, &kgsl_driver.stats.page_alloc); } for_each_sg(memdesc->sgt->sgl, sg, memdesc->sgt->nents, i) { /* * sg_alloc_table_from_pages() will collapse any physically * adjacent pages into a single scatterlist entry. We cannot * just call __free_pages() on the entire set since we cannot * ensure that the size is a whole order. Instead, free each * page or compound page group individually. */ struct page *p = sg_page(sg), *next; unsigned int j = 0, count; while (j < (sg->length/PAGE_SIZE)) { if (memdesc->priv & KGSL_MEMDESC_TZ_LOCKED) ClearPagePrivate(p); count = 1 << compound_order(p); next = nth_page(p, count); __free_pages(p, compound_order(p)); p = next; j += count; } } } /* * kgsl_page_alloc_map_kernel - Map the memory in memdesc to kernel address * space * * @memdesc - The memory descriptor which contains information about the memory * * Return: 0 on success else error code */ static int kgsl_page_alloc_map_kernel(struct kgsl_memdesc *memdesc) { int ret = 0; /* Sanity check - don't map more than we could possibly chew */ if (memdesc->size > ULONG_MAX) return -ENOMEM; mutex_lock(&kernel_map_global_lock); if (!memdesc->hostptr) { pgprot_t page_prot = pgprot_writecombine(PAGE_KERNEL); struct page **pages = NULL; struct scatterlist *sg; int npages = PAGE_ALIGN(memdesc->size) >> PAGE_SHIFT; int sglen = memdesc->sgt->nents; int i, count = 0; /* create a list of pages to call vmap */ pages = kgsl_malloc(npages * sizeof(struct page *)); if (pages == NULL) { ret = -ENOMEM; goto done; } for_each_sg(memdesc->sgt->sgl, sg, sglen, i) { struct page *page = sg_page(sg); int j; for (j = 0; j < sg->length >> PAGE_SHIFT; j++) pages[count++] = page++; } memdesc->hostptr = vmap(pages, count, VM_IOREMAP, page_prot); if (memdesc->hostptr) KGSL_STATS_ADD(memdesc->size, &kgsl_driver.stats.vmalloc, &kgsl_driver.stats.vmalloc_max); else ret = -ENOMEM; kgsl_free(pages); } if (memdesc->hostptr) memdesc->hostptr_count++; done: mutex_unlock(&kernel_map_global_lock); return ret; } static int kgsl_contiguous_vmfault(struct kgsl_memdesc *memdesc, struct vm_area_struct *vma, struct vm_fault *vmf) { unsigned long offset, pfn; int ret; offset = ((unsigned long) vmf->virtual_address - vma->vm_start) >> PAGE_SHIFT; pfn = (memdesc->physaddr >> PAGE_SHIFT) + offset; ret = vm_insert_pfn(vma, (unsigned long) vmf->virtual_address, pfn); if (ret == -ENOMEM || ret == -EAGAIN) return VM_FAULT_OOM; else if (ret == -EFAULT) return VM_FAULT_SIGBUS; return VM_FAULT_NOPAGE; } static void kgsl_cma_coherent_free(struct kgsl_memdesc *memdesc) { struct dma_attrs *attrs = NULL; if (memdesc->hostptr) { if (memdesc->priv & KGSL_MEMDESC_SECURE) { atomic_long_sub(memdesc->size, &kgsl_driver.stats.secure); kgsl_cma_unlock_secure(memdesc); attrs = &memdesc->attrs; } else atomic_long_sub(memdesc->size, &kgsl_driver.stats.coherent); dma_free_attrs(memdesc->dev, (size_t) memdesc->size, memdesc->hostptr, memdesc->physaddr, attrs); } } /* Global */ static struct kgsl_memdesc_ops kgsl_page_alloc_ops = { .free = kgsl_page_alloc_free, .vmflags = VM_DONTDUMP | VM_DONTEXPAND | VM_DONTCOPY, .vmfault = kgsl_page_alloc_vmfault, .map_kernel = kgsl_page_alloc_map_kernel, .unmap_kernel = kgsl_page_alloc_unmap_kernel, }; /* CMA ops - used during NOMMU mode */ static struct kgsl_memdesc_ops kgsl_cma_ops = { .free = kgsl_cma_coherent_free, .vmflags = VM_DONTDUMP | VM_PFNMAP | VM_DONTEXPAND | VM_DONTCOPY, .vmfault = kgsl_contiguous_vmfault, }; #ifdef CONFIG_ARM64 /* * For security reasons, ARMv8 doesn't allow invalidate only on read-only * mapping. It would be performance prohibitive to read the permissions on * the buffer before the operation. Every use case that we have found does not * assume that an invalidate operation is invalidate only, so we feel * comfortable turning invalidates into flushes for these targets */ static inline unsigned int _fixup_cache_range_op(unsigned int op) { if (op == KGSL_CACHE_OP_INV) return KGSL_CACHE_OP_FLUSH; return op; } #else static inline unsigned int _fixup_cache_range_op(unsigned int op) { return op; } #endif int kgsl_cache_range_op(struct kgsl_memdesc *memdesc, uint64_t offset, uint64_t size, unsigned int op) { /* * If the buffer is mapped in the kernel operate on that address * otherwise use the user address */ void *addr = (memdesc->hostptr) ? memdesc->hostptr : (void *) memdesc->useraddr; /* Make sure that size is non-zero */ if (!size) return -EINVAL; /* Make sure that the offset + size isn't bigger than we can handle */ if ((offset + size) > ULONG_MAX) return -ERANGE; /* Make sure the offset + size do not overflow the address */ if (addr + ((size_t) offset + (size_t) size) < addr) return -ERANGE; /* Check that offset+length does not exceed memdesc->size */ if (offset + size > memdesc->size) return -ERANGE; /* Return quietly if the buffer isn't mapped on the CPU */ if (addr == NULL) return 0; addr = addr + offset; /* * The dmac_xxx_range functions handle addresses and sizes that * are not aligned to the cacheline size correctly. */ switch (_fixup_cache_range_op(op)) { case KGSL_CACHE_OP_FLUSH: dmac_flush_range(addr, addr + (size_t) size); break; case KGSL_CACHE_OP_CLEAN: dmac_clean_range(addr, addr + (size_t) size); break; case KGSL_CACHE_OP_INV: dmac_inv_range(addr, addr + (size_t) size); break; } return 0; } EXPORT_SYMBOL(kgsl_cache_range_op); #ifndef CONFIG_ALLOC_BUFFERS_IN_4K_CHUNKS static inline int get_page_size(size_t size, unsigned int align) { return (align >= ilog2(SZ_64K) && size >= SZ_64K) ? SZ_64K : PAGE_SIZE; } #else static inline int get_page_size(size_t size, unsigned int align) { return PAGE_SIZE; } #endif static int _kgsl_sharedmem_page_alloc(struct kgsl_memdesc *memdesc, struct kgsl_pagetable *pagetable, uint64_t size) { int ret = 0; unsigned int j, pcount = 0, page_size, len_alloc; size_t len; struct page **pages = NULL; pgprot_t page_prot = pgprot_writecombine(PAGE_KERNEL); void *ptr; unsigned int align; unsigned int step = ((VMALLOC_END - VMALLOC_START)/8) >> PAGE_SHIFT; align = (memdesc->flags & KGSL_MEMALIGN_MASK) >> KGSL_MEMALIGN_SHIFT; page_size = get_page_size(size, align); /* * The alignment cannot be less than the intended page size - it can be * larger however to accomodate hardware quirks */ if (align < ilog2(page_size)) kgsl_memdesc_set_align(memdesc, ilog2(page_size)); if (size > SIZE_MAX) return -EINVAL; /* * There needs to be enough room in the page array to be able to * service the allocation entirely with PAGE_SIZE sized chunks */ len_alloc = PAGE_ALIGN(size) >> PAGE_SHIFT; memdesc->pagetable = pagetable; memdesc->ops = &kgsl_page_alloc_ops; memdesc->sgt = kmalloc(sizeof(struct sg_table), GFP_KERNEL); if (memdesc->sgt == NULL) return -ENOMEM; /* * Allocate space to store the list of pages to send to vmap. This is an * array of pointers so we can track 1024 pages per page of allocation */ pages = kgsl_malloc(len_alloc * sizeof(struct page *)); if (pages == NULL) { ret = -ENOMEM; goto done; } len = size; while (len > 0) { struct page *page; gfp_t gfp_mask = __GFP_HIGHMEM; int j; /* don't waste space at the end of the allocation*/ if (len < page_size) page_size = PAGE_SIZE; /* * Don't do some of the more aggressive memory recovery * techniques for large order allocations */ if (page_size != PAGE_SIZE) gfp_mask |= __GFP_COMP | __GFP_NORETRY | __GFP_NO_KSWAPD | __GFP_NOWARN; else gfp_mask |= GFP_KERNEL; if (sharedmem_noretry_flag == true) gfp_mask |= __GFP_NORETRY | __GFP_NOWARN; page = alloc_pages(gfp_mask, get_order(page_size)); if (page == NULL) { if (page_size != PAGE_SIZE) { page_size = PAGE_SIZE; continue; } /* * Update sglen and memdesc size,as requested allocation * not served fully. So that they can be correctly freed * in kgsl_sharedmem_free(). */ memdesc->size = (size - len); if (sharedmem_noretry_flag != true) KGSL_CORE_ERR( "Out of memory: only allocated %lldKB of %lldKB requested\n", (size - len) >> 10, size >> 10); ret = -ENOMEM; goto done; } for (j = 0; j < page_size >> PAGE_SHIFT; j++) pages[pcount++] = nth_page(page, j); len -= page_size; memdesc->size += page_size; } ret = sg_alloc_table_from_pages(memdesc->sgt, pages, pcount, 0, memdesc->size, GFP_KERNEL); if (ret) goto done; /* Call to the hypervisor to lock any secure buffer allocations */ if (memdesc->flags & KGSL_MEMFLAGS_SECURE) { unsigned int i; struct scatterlist *sg; int dest_perms = PERM_READ | PERM_WRITE; int source_vm = VMID_HLOS; int dest_vm = VMID_CP_PIXEL; ret = hyp_assign_table(memdesc->sgt, &source_vm, 1, &dest_vm, &dest_perms, 1); if (ret) goto done; /* Set private bit for each sg to indicate that its secured */ for_each_sg(memdesc->sgt->sgl, sg, memdesc->sgt->nents, i) SetPagePrivate(sg_page(sg)); memdesc->priv |= KGSL_MEMDESC_TZ_LOCKED; /* Record statistics */ KGSL_STATS_ADD(memdesc->size, &kgsl_driver.stats.secure, &kgsl_driver.stats.secure_max); /* Don't map and zero the locked secure buffer */ goto done; } /* * All memory that goes to the user has to be zeroed out before it gets * exposed to userspace. This means that the memory has to be mapped in * the kernel, zeroed (memset) and then unmapped. This also means that * the dcache has to be flushed to ensure coherency between the kernel * and user pages. We used to pass __GFP_ZERO to alloc_page which mapped * zeroed and unmaped each individual page, and then we had to turn * around and call flush_dcache_page() on that page to clear the caches. * This was killing us for performance. Instead, we found it is much * faster to allocate the pages without GFP_ZERO, map a chunk of the * range ('step' pages), memset it, flush it and then unmap * - this results in a factor of 4 improvement for speed for large * buffers. There is a small decrease in speed for small buffers, * but only on the order of a few microseconds at best. The 'step' * size is based on a guess at the amount of free vmalloc space, but * will scale down if there's not enough free space. */ for (j = 0; j < pcount; j += step) { step = min(step, pcount - j); ptr = vmap(&pages[j], step, VM_IOREMAP, page_prot); if (ptr != NULL) { memset(ptr, 0, step * PAGE_SIZE); dmac_flush_range(ptr, ptr + step * PAGE_SIZE); vunmap(ptr); } else { int k; /* Very, very, very slow path */ for (k = j; k < j + step; k++) { ptr = kmap_atomic(pages[k]); memset(ptr, 0, PAGE_SIZE); dmac_flush_range(ptr, ptr + PAGE_SIZE); kunmap_atomic(ptr); } /* scale down the step size to avoid this path */ if (step > 1) step >>= 1; } } KGSL_STATS_ADD(memdesc->size, &kgsl_driver.stats.page_alloc, &kgsl_driver.stats.page_alloc_max); done: if (ret) { unsigned int count = 1; for (j = 0; j < pcount; j += count) { count = 1 << compound_order(pages[j]); __free_pages(pages[j], compound_order(pages[j])); } kfree(memdesc->sgt); memset(memdesc, 0, sizeof(*memdesc)); } kgsl_free(pages); return ret; } int kgsl_sharedmem_page_alloc_user(struct kgsl_memdesc *memdesc, struct kgsl_pagetable *pagetable, uint64_t size) { size = PAGE_ALIGN(size); if (size == 0) return -EINVAL; return _kgsl_sharedmem_page_alloc(memdesc, pagetable, size); } EXPORT_SYMBOL(kgsl_sharedmem_page_alloc_user); void kgsl_sharedmem_free(struct kgsl_memdesc *memdesc) { if (memdesc == NULL || memdesc->size == 0) return; if (memdesc->gpuaddr) { kgsl_mmu_unmap(memdesc->pagetable, memdesc); kgsl_mmu_put_gpuaddr(memdesc->pagetable, memdesc); } if (memdesc->ops && memdesc->ops->free) memdesc->ops->free(memdesc); if (memdesc->sgt) { sg_free_table(memdesc->sgt); kfree(memdesc->sgt); } memset(memdesc, 0, sizeof(*memdesc)); } EXPORT_SYMBOL(kgsl_sharedmem_free); int kgsl_sharedmem_readl(const struct kgsl_memdesc *memdesc, uint32_t *dst, uint64_t offsetbytes) { uint32_t *src; BUG_ON(memdesc == NULL || memdesc->hostptr == NULL || dst == NULL); WARN_ON(offsetbytes % sizeof(uint32_t) != 0); if (offsetbytes % sizeof(uint32_t) != 0) return -EINVAL; WARN_ON(offsetbytes + sizeof(uint32_t) > memdesc->size); if (offsetbytes + sizeof(uint32_t) > memdesc->size) return -ERANGE; rmb(); src = (uint32_t *)(memdesc->hostptr + offsetbytes); *dst = *src; return 0; } EXPORT_SYMBOL(kgsl_sharedmem_readl); int kgsl_sharedmem_writel(struct kgsl_device *device, const struct kgsl_memdesc *memdesc, uint64_t offsetbytes, uint32_t src) { uint32_t *dst; BUG_ON(memdesc == NULL || memdesc->hostptr == NULL); WARN_ON(offsetbytes % sizeof(uint32_t) != 0); if (offsetbytes % sizeof(uint32_t) != 0) return -EINVAL; WARN_ON(offsetbytes + sizeof(uint32_t) > memdesc->size); if (offsetbytes + sizeof(uint32_t) > memdesc->size) return -ERANGE; kgsl_cffdump_write(device, memdesc->gpuaddr + offsetbytes, src); dst = (uint32_t *)(memdesc->hostptr + offsetbytes); *dst = src; wmb(); return 0; } EXPORT_SYMBOL(kgsl_sharedmem_writel); int kgsl_sharedmem_readq(const struct kgsl_memdesc *memdesc, uint64_t *dst, uint64_t offsetbytes) { uint64_t *src; BUG_ON(memdesc == NULL || memdesc->hostptr == NULL || dst == NULL); WARN_ON(offsetbytes % sizeof(uint32_t) != 0); if (offsetbytes % sizeof(uint32_t) != 0) return -EINVAL; WARN_ON(offsetbytes + sizeof(uint32_t) > memdesc->size); if (offsetbytes + sizeof(uint32_t) > memdesc->size) return -ERANGE; /* * We are reading shared memory between CPU and GPU. * Make sure reads before this are complete */ rmb(); src = (uint64_t *)(memdesc->hostptr + offsetbytes); *dst = *src; return 0; } EXPORT_SYMBOL(kgsl_sharedmem_readq); int kgsl_sharedmem_writeq(struct kgsl_device *device, const struct kgsl_memdesc *memdesc, uint64_t offsetbytes, uint64_t src) { uint64_t *dst; BUG_ON(memdesc == NULL || memdesc->hostptr == NULL); WARN_ON(offsetbytes % sizeof(uint32_t) != 0); if (offsetbytes % sizeof(uint32_t) != 0) return -EINVAL; WARN_ON(offsetbytes + sizeof(uint32_t) > memdesc->size); if (offsetbytes + sizeof(uint32_t) > memdesc->size) return -ERANGE; kgsl_cffdump_write(device, lower_32_bits(memdesc->gpuaddr + offsetbytes), src); kgsl_cffdump_write(device, upper_32_bits(memdesc->gpuaddr + offsetbytes), src); dst = (uint64_t *)(memdesc->hostptr + offsetbytes); *dst = src; /* * We are writing to shared memory between CPU and GPU. * Make sure write above is posted immediately */ wmb(); return 0; } EXPORT_SYMBOL(kgsl_sharedmem_writeq); int kgsl_sharedmem_set(struct kgsl_device *device, const struct kgsl_memdesc *memdesc, uint64_t offsetbytes, unsigned int value, uint64_t sizebytes) { BUG_ON(memdesc == NULL || memdesc->hostptr == NULL); BUG_ON(offsetbytes + sizebytes > memdesc->size); kgsl_cffdump_memset(device, memdesc->gpuaddr + offsetbytes, value, sizebytes); memset(memdesc->hostptr + offsetbytes, value, sizebytes); return 0; } EXPORT_SYMBOL(kgsl_sharedmem_set); static const char * const memtype_str[] = { [KGSL_MEMTYPE_OBJECTANY] = "any(0)", [KGSL_MEMTYPE_FRAMEBUFFER] = "framebuffer", [KGSL_MEMTYPE_RENDERBUFFER] = "renderbuffer", [KGSL_MEMTYPE_ARRAYBUFFER] = "arraybuffer", [KGSL_MEMTYPE_ELEMENTARRAYBUFFER] = "elementarraybuffer", [KGSL_MEMTYPE_VERTEXARRAYBUFFER] = "vertexarraybuffer", [KGSL_MEMTYPE_TEXTURE] = "texture", [KGSL_MEMTYPE_SURFACE] = "surface", [KGSL_MEMTYPE_EGL_SURFACE] = "egl_surface", [KGSL_MEMTYPE_GL] = "gl", [KGSL_MEMTYPE_CL] = "cl", [KGSL_MEMTYPE_CL_BUFFER_MAP] = "cl_buffer_map", [KGSL_MEMTYPE_CL_BUFFER_NOMAP] = "cl_buffer_nomap", [KGSL_MEMTYPE_CL_IMAGE_MAP] = "cl_image_map", [KGSL_MEMTYPE_CL_IMAGE_NOMAP] = "cl_image_nomap", [KGSL_MEMTYPE_CL_KERNEL_STACK] = "cl_kernel_stack", [KGSL_MEMTYPE_COMMAND] = "command", [KGSL_MEMTYPE_2D] = "2d", [KGSL_MEMTYPE_EGL_IMAGE] = "egl_image", [KGSL_MEMTYPE_EGL_SHADOW] = "egl_shadow", [KGSL_MEMTYPE_MULTISAMPLE] = "egl_multisample", /* KGSL_MEMTYPE_KERNEL handled below, to avoid huge array */ }; void kgsl_get_memory_usage(char *name, size_t name_size, uint64_t memflags) { unsigned int type = MEMFLAGS(memflags, KGSL_MEMTYPE_MASK, KGSL_MEMTYPE_SHIFT); if (type == KGSL_MEMTYPE_KERNEL) strlcpy(name, "kernel", name_size); else if (type < ARRAY_SIZE(memtype_str) && memtype_str[type] != NULL) strlcpy(name, memtype_str[type], name_size); else snprintf(name, name_size, "unknown(%3d)", type); } EXPORT_SYMBOL(kgsl_get_memory_usage); int kgsl_cma_alloc_coherent(struct kgsl_device *device, struct kgsl_memdesc *memdesc, struct kgsl_pagetable *pagetable, uint64_t size) { int result = 0; size = ALIGN(size, PAGE_SIZE); if (size == 0 || size > SIZE_MAX) return -EINVAL; memdesc->size = size; memdesc->pagetable = pagetable; memdesc->ops = &kgsl_cma_ops; memdesc->dev = device->dev->parent; memdesc->hostptr = dma_alloc_attrs(memdesc->dev, (size_t) size, &memdesc->physaddr, GFP_KERNEL, NULL); if (memdesc->hostptr == NULL) { result = -ENOMEM; goto err; } result = memdesc_sg_dma(memdesc, memdesc->physaddr, size); if (result) goto err; /* Record statistics */ KGSL_STATS_ADD(size, &kgsl_driver.stats.coherent, &kgsl_driver.stats.coherent_max); err: if (result) kgsl_sharedmem_free(memdesc); return result; } EXPORT_SYMBOL(kgsl_cma_alloc_coherent); static int scm_lock_chunk(struct kgsl_memdesc *memdesc, int lock) { struct cp2_lock_req request; unsigned int resp; unsigned int *chunk_list; struct scm_desc desc = {0}; int result; /* * Flush the virt addr range before sending the memory to the * secure environment to ensure the data is actually present * in RAM * * Chunk_list holds the physical address of secure memory. * Pass in the virtual address of chunk_list to flush. * Chunk_list size is 1 because secure memory is physically * contiguous. */ chunk_list = kzalloc(sizeof(unsigned int), GFP_KERNEL); if (!chunk_list) return -ENOMEM; chunk_list[0] = memdesc->physaddr; dmac_flush_range((void *)chunk_list, (void *)chunk_list + 1); request.chunks.chunk_list = virt_to_phys(chunk_list); /* * virt_to_phys(chunk_list) may be an address > 4GB. It is guaranteed * that when using scm_call (the older interface), the phys addresses * will be restricted to below 4GB. */ desc.args[0] = virt_to_phys(chunk_list); desc.args[1] = request.chunks.chunk_list_size = 1; desc.args[2] = request.chunks.chunk_size = (unsigned int) memdesc->size; desc.args[3] = request.mem_usage = 0; desc.args[4] = request.lock = lock; desc.args[5] = 0; desc.arginfo = SCM_ARGS(6, SCM_RW, SCM_VAL, SCM_VAL, SCM_VAL, SCM_VAL, SCM_VAL); kmap_flush_unused(); kmap_atomic_flush_unused(); if (!is_scm_armv8()) { result = scm_call(SCM_SVC_MP, MEM_PROTECT_LOCK_ID2, &request, sizeof(request), &resp, sizeof(resp)); } else { result = scm_call2(SCM_SIP_FNID(SCM_SVC_MP, MEM_PROTECT_LOCK_ID2_FLAT), &desc); resp = desc.ret[0]; } kfree(chunk_list); return result; } int kgsl_cma_alloc_secure(struct kgsl_device *device, struct kgsl_memdesc *memdesc, uint64_t size) { struct kgsl_iommu *iommu = device->mmu.priv; int result = 0; struct kgsl_pagetable *pagetable = device->mmu.securepagetable; size_t aligned; if (size == 0) return -EINVAL; /* Align size to 1M boundaries */ aligned = ALIGN(size, SZ_1M); /* The SCM call uses an unsigned int for the size */ if (aligned > UINT_MAX) return -EINVAL; /* * If there is more than a page gap between the requested size and the * aligned size we don't need to add more memory for a guard page. Yay! */ if (memdesc->priv & KGSL_MEMDESC_GUARD_PAGE) if (aligned - size >= SZ_4K) memdesc->priv &= ~KGSL_MEMDESC_GUARD_PAGE; memdesc->size = aligned; memdesc->pagetable = pagetable; memdesc->ops = &kgsl_cma_ops; memdesc->dev = iommu->ctx[KGSL_IOMMU_CONTEXT_SECURE].dev; init_dma_attrs(&memdesc->attrs); dma_set_attr(DMA_ATTR_STRONGLY_ORDERED, &memdesc->attrs); memdesc->hostptr = dma_alloc_attrs(memdesc->dev, aligned, &memdesc->physaddr, GFP_KERNEL, &memdesc->attrs); if (memdesc->hostptr == NULL) { result = -ENOMEM; goto err; } result = memdesc_sg_dma(memdesc, memdesc->physaddr, aligned); if (result) goto err; result = scm_lock_chunk(memdesc, 1); if (result != 0) goto err; /* Set the private bit to indicate that we've secured this */ SetPagePrivate(sg_page(memdesc->sgt->sgl)); memdesc->priv |= KGSL_MEMDESC_TZ_LOCKED; /* Record statistics */ KGSL_STATS_ADD(aligned, &kgsl_driver.stats.secure, &kgsl_driver.stats.secure_max); err: if (result) kgsl_sharedmem_free(memdesc); return result; } EXPORT_SYMBOL(kgsl_cma_alloc_secure); /** * kgsl_cma_unlock_secure() - Unlock secure memory by calling TZ * @memdesc: memory descriptor */ static void kgsl_cma_unlock_secure(struct kgsl_memdesc *memdesc) { if (memdesc->size == 0 || !(memdesc->priv & KGSL_MEMDESC_TZ_LOCKED)) return; if (!scm_lock_chunk(memdesc, 0)) ClearPagePrivate(sg_page(memdesc->sgt->sgl)); } void kgsl_sharedmem_set_noretry(bool val) { sharedmem_noretry_flag = val; } bool kgsl_sharedmem_get_noretry(void) { return sharedmem_noretry_flag; }