- 14 Dec, 2013 40 commits
-
-
Colin Cross authored
Add ion_heap_pages_zero for ion heaps to use to zero pages during initialization or allocation, when a struct ion_buffer may not be available. Use it from the chunk heap and carveout heaps. Signed-off-by:
Colin Cross <ccross@android.com> Signed-off-by:
John Stultz <john.stultz@linaro.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Colin Cross authored
The carveout heap wasn't zeroing its buffers after use. Create the sg_table during allocate instead of map_dma, to allow using the sg_table during free, and call ion_heap_buffer_zero during free. Also fixes a missing kfree when destroying the table. Signed-off-by:
-
Colin Cross authored
Fix sparse warnings in ion. Signed-off-by:
-
Colin Cross authored
There is no reason to use kzalloc, just call alloc_pages directly. Change the GFP from GFP_KERNEL to include __GFP_HIGH, to allow it to return contiguous pages from highmem. virt_to_* functions aren't valid on highmem pages, so store the struct page * in an sg_table in buffer->priv_virt like most other heaps, and replace virt_to_* with page_to_*. Signed-off-by:
-
Colin Cross authored
Now that ion_vm_fault uses vm_insert_pfn instead of vm_insert_page cached buffers can be supported in any heap. Remove the checks in the chunk and system heaps. Signed-off-by:
-
Colin Cross authored
Now that ion_vm_fault doesn't need a struct page with a nonzero refcount, there is no need allocate heap memory for cached pages using split_page. Remove the ion_heap_alloc_pages and ion_heap_free_pages helpers in favor of direct calls to alloc_pages and __free_pages, and remove the special handling in the system heap. Signed-off-by:
-
Colin Cross authored
Most ion userspace mappings are created with remap_pfn_range. Use vm_insert_pfn instead of vm_insert_page to make faulted cached mappings look more like uncached mappings. Signed-off-by:
-
Colin Cross authored
Check the return value of remap_pfn_range and return an error if it fails. Signed-off-by:
-
Colin Cross authored
When the shrinkers are called with GFP_HIGH free low memory first, it is more important to have free than high memory. Signed-off-by:
-
Colin Cross authored
ion_heap_buffer_zero can spend a long time in unmap_kernel_range if it has to broadcast a tlb flush to every cpu for every page. Modify it to batch pages into a larger region to clear using a single mapping. This may cause the mapping size to change if the buffer size is not a multiple of the mapping size, so switch to allocating the address space for each chunk. This allows us to use vm_map_ram to handle the allocation and mapping together. The number of pages to zero using a single mapping is set to 32 to hit the fastpath in vm_map_ram. Signed-off-by:
-
Colin Cross authored
Signed-off-by:
-
Colin Cross authored
Ion will compile and run on other platforms now, remove the dependency on ARM. Signed-off-by:
-
John Stultz authored
In testing ion system heap allocations, I ran across two issues: 1) Not k*z*allocing the sg table. This can cause trouble if we end up trying call sg_alloc_table() with too many entries, then sg_alloc_table() internally fails and tries to free what it thinks is internal table structure, which causes bad pointer traversals. 2) The second list_for_each_entry probably should be _safe, since I was seeing strange lock warnings and oopses on occasion. This seems to resolve it, but could use some extra checking. Signed-off-by:
-
Mitchel Humpherys authored
The `buffer' variable is being used after being freed. Fix this. Signed-off-by:
Mitchel Humpherys <mitchelh@codeaurora.org> Signed-off-by:
-
Colin Cross authored
Convert the ion ioctls to use _IOW instead of _IOWR where appropriate, and factor out the copy_from_user and copy_to_user based on the _IOC_DIR bits. For the existing incorrect ioctls, add a function to wrap _IOC_DIR to return the corrected value. Signed-off-by:
-
Colin Cross authored
ion_system_contig_heap buffers have an sglist, just call ion_heap_map_user to map it. Signed-off-by:
-
Colin Cross authored
Use %z for size_t and %pa for dma_addr_t to avoid warnings in printks. Signed-off-by:
-
Colin Cross authored
phys_to_page and __phys_to_pfn don't exist on all platforms. Use a combination of pfn_to_page, PFN_DOWN, page_to_pfn, and virt_to_page to get the same results. Signed-off-by:
-
Colin Cross authored
ion_heap_map_kernel already implements mapping a scatterlist of pages into the kernel, and all heaps are required to have struct pages associated with them, so delete the functions that use __arm_ioremap and use ion_heap_map_kernel instead. Signed-off-by:
-
Colin Cross authored
Use idr_alloc instead if idr_pre_get/idr_get_new_above, and remove idr_remove_all. Signed-off-by:
-
Colin Cross authored
Add a /dev/ion-test device that will be created if CONFIG_ION_TEST is set. The device accepts a dma_buf fd and allows reading and writing to the backing memory using DMA-like apis or kernel mapping apis. Can be used to test the dma_buf mapping ops, including the ion implementations, from userspace. Signed-off-by:
-
Colin Cross authored
ion_system_heap can only satisfy page alignment, and ion_system_contig_heap can only satisify alignment to the allocation size. Neither can support faulting user mappings because they use slab pages. Signed-off-by:
-
Colin Cross authored
ion is always dealing with the allocation and not the mapping, so it should always be using sg->length and not sg->dma_length. Signed-off-by:
-
Colin Cross authored
__dma_page_cpu_to_dev is a private ARM api that is not available on 3.10 and was never available on other architectures. We can get the same behavior by calling dma_sync_sg_for_device with a scatterlist containing a single page. It's still not quite a kosher use of the dma apis, we still conflate physical addresses with bus addresses, but it should at least compile on all platforms, and work on any platform that doesn't have a physical to bus address translation. Signed-off-by:
-
Colin Cross authored
If userspace passes a length between -4095 and -1 to allocate it will pass the len != 0 check, but when len is page aligned it will be 0. Check len after page aligning. Drop the warning as well, userspace shouldn't be able to trigger a warning in the kernel. Signed-off-by:
-
Colin Cross authored
commit 1262ab1846cf76f7549c66ef709120dbfbe6d49f (ion: replace userspace handle cookies with idr) broke the locking in ion. The ION_IOC_FREE and ION_IOC_MAP ioctls were relying on ion_handle_validate to detect the case where a call raced with another ION_IOC_FREE which may have freed the struct ion_handle. Rename ion_uhandle_get to ion_handle_get_by_id, and have it take the client lock and return with an extra reference to the handle. Make each caller put its reference once it is done with the handle. Also modify users of ion_handle_validate to continue to hold the client lock after calling ion_handle_validate until they are done with the handle, and warn if ion_handle_validate is called without the client lock held. Signed-off-by:
-
John Stultz authored
The compat support added to ion didn't provide compat ioctl numbers (who's value depends on the compat structure size). So 32bit applications would get an error when trying to make ioctl calls. This patch adds the needed COMPAT_ macros and uses them in the compat_ion_ioctl, translating them to their non-compat cmd when calling the normal ioctl call. Signed-off-by:
-
Colin Cross authored
Split the userspace api out of drivers/staging/android/ion/ion.h into drivers/staging/android/uapi/ion.h Signed-off-by:
-
Colin Cross authored
The mapper abstraction layer was removed before the initial ion commit, but a stray ion_system_mapper.c file was left in. Delete it. Signed-off-by:
-
Rom Lemarchand authored
Add a compat_ioctl to the ion driver Signed-off-by:
Rom Lemarchand <romlem@google.com> [jstultz: modified patch to apply to staging directory] Signed-off-by:
-
Rom Lemarchand authored
Turn ion_user_handle_t to int. This change reflects the underlying type returned by the ion driver. Signed-off-by:
-
Rom Lemarchand authored
Declare new ion_user_handle_t type to contain the token returned to user-space. This allows a 2-step migration of the user-space code to a new kernel header first, then will allow us to change the definition of the ion_user_handle_type_t to int without breaking the API. Signed-off-by:
-
Colin Cross authored
ion userspace clients think that the cookie is a pointer, so they use NULL to check if the handle has been initialized. Set the first id number to 1. Signed-off-by:
-
Colin Cross authored
The only remaining users of the client->handles rbtree are iterating through it like a list. Keep the rbtree, but change its index to be the buffer address instead of the handle address, which makes ion_handle_lookup a fast rbtree search. Signed-off-by:
-
Colin Cross authored
Userspace handles should not leak kernel virtual addresses to userspace. They have to be validated by looking them up in an rbtree anyways, so replace them with an idr and validate them by using idr_find to convert the id number to the struct ion_handle pointer. Signed-off-by:
-
Colin Cross authored
IS_ERR_OR_NULL is often part of a bad pattern that can accidentally return 0 on error: if (IS_ERR_OR_NULL(ptr)) return PTR_ERR(ptr); It also usually means that the errors of a function are not well defined. Replace all uses in ion.c by ensure that the return type of any function in ion is an ERR_PTR. Specify that the expected return value from map_kernel or map_dma heap ops is ERR_PTR, and warn if a heap returns NULL. Signed-off-by: -
Colin Cross authored
ion is going to stop accepting NULL as an error value, use ERR_PTR. Signed-off-by:
-
Colin Cross authored
Signed-off-by:
-
Colin Cross authored
buffer->size is controlled by the outer ion layer, don't modify it inside the heap. Instead, compute the rounded up allocated size on demand. Signed-off-by:
-
Rebecca Schultz Zavin authored
Previously the code to fault ion buffers in one page at a time had a performance problem caused by the requirement to traverse the sg list looking for the right page to load in (a result of the fact that the items in the list may not be of uniform size). To fix the problem, for buffers that will be faulted in, also keep a flat array of all the pages in the buffer to use from the fault handler. To recover some of the additional memory footprint this creates per buffer, dirty bits used to indicate which pages have been faulted in to the cpu are now stored in the low bit of each page struct pointer in the page array. Signed-off-by:
Rebecca Schultz Zavin <rebecca@android.com> [jstultz: modified patch to apply to staging directory] Signed-off-by:
-
