summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2006-10-26New mm function names. Update header.Thomas Hellstrom
2006-10-26Add improved alignment functionality to the core memory manager.Thomas Hellstrom
This makes an allocated block actually align itself and returns any wasted space to the manager. Also add some functions to grow and shrink the managed area. This will be used in the future to manage the buffer object swap cache.
2006-10-26Add a one-page hole in the file offset space between buffers.Thomas Hellstrom
2006-10-21Merge branch 'master' of git+ssh://git.freedesktop.org/git/mesa/drmThomas Hellstrom
2006-10-21The CPU cache must be flushed _before_ we start modifying the kernel map ptes,Thomas Hellstrom
otherwise data will be missing, which becomes apparent when the kernel evicts batch buffers which are likely to be written into in the evicted state, and then rebound to the AGP aperture. This means we cannot rely on the AGP module to flush the cache for us.
2006-10-20Bug #1746: Set dev_priv_size for the MGA driver.Tilman Sauerbeck
2006-10-20We apparently need this global cache flush anyway.Thomas Hellstrom
2006-10-20Bug #8707, 2.6.19-rc compatibility for memory manager code.Thomas Hellstrom
2006-10-19Merge branch 'master' of git+ssh://git.freedesktop.org/git/mesa/drmThomas Hellstrom
2006-10-19Importing fixes from drm-ttm-0-2-branchThomas Hellstrom
2006-10-19Make sure delayed delete list is empty on lastclose.Thomas Hellstrom
Fix some refcounting errors. Fix some error messages.
2006-10-18Merging drm-ttm-0-2-branchThomas Hellstrom
Conflicts: linux-core/drmP.h linux-core/drm_drv.c linux-core/drm_irq.c linux-core/drm_stub.c shared-core/drm.h shared-core/i915_drv.h shared-core/i915_irq.c
2006-10-18Remove stray softlink.Thomas Hellstrom
2006-10-18Avoid driver-specific AGP user-populated types, since we don't know what AGPThomas Hellstrom
driver we're on. Avoid global cache flushes before inserting pages. In general, they are never mapped, and not accessed through the kernel map, so a cache flush should not be necessary. The exception is pages that are bound cached. We might need a cache flush for those.
2006-10-17Remove some debugging messages.Thomas Hellstrom
2006-10-17Remove max number of locked pages check and call, sinceThomas Hellstrom
that is now handled by the memory accounting.
2006-10-17Lindent.Thomas Hellstrom
2006-10-17Add vma list memory usage to memory accounting.Thomas Hellstrom
Use byte unit for /proc printout of memory usage for small sizes to be able to detect memory allocation bugs more easily.
2006-10-17Add memory usage accounting to avoid DOS problems.Thomas Hellstrom
2006-10-17Implement mm_lock and mm_unlock functions.Thomas Hellstrom
The mm_lock function is used when leaving vt. It evicts _all_ buffers. Buffers with the DRM_BO_NO_MOVE attribute set will be guaranteed to get the same offset when / if they are rebound.
2006-10-17Remove the memory manager parameter from the put_block function, as thisThomas Hellstrom
makes the client code a lot cleaner. Prepare buffer manager for lock and unlock calls.
2006-10-17Extend generality for more memory types.Thomas Hellstrom
Fix up init and destruction code.
2006-10-16dev->agp_buffer_map is not initialized for AGP DMA on savagesMichael Karcher
bug 8662
2006-10-16Change Intel AGP memory type numbers.Thomas Hellstrom
2006-10-14remove config.h from build no longer exists kbuild does itDave Airlie
2006-10-12Bugfixes.Thomas Hellstrom
2006-10-12Simplify the AGP backend interface somewhat.Thomas Hellstrom
Fix buffer bound caching policy changing, Allow on-the-fly changing of caching policy on bound buffers if the hardware supports it. Allow drivers to use driver-specific AGP memory types for TTM AGP pages. Will make AGP drivers much easier to migrate.
2006-10-11Compatibility code for 2.6.15-2.6.18. It is ugly but a little comfort is thatThomas Hellstrom
it will go away in the mainstream kernel. Some bugfixes, mainly in error paths.
2006-10-11Big update:Thomas Hellstrom
Adapt for new functions in the 2.6.19 kernel. Remove the ability to have multiple regions in one TTM. This simplifies a lot of code. Remove the ability to access TTMs from user space. We don't need it anymore without ttm regions. Don't change caching policy for evicted buffers. Instead change it only when the buffer is accessed by the CPU (on the first page fault). This tremendously speeds up eviction rates. Current code is safe for kernels <= 2.6.14. Should also be OK with 2.6.19 and above.
2006-10-10Use a nopage-based approach to fault in pfns.Thomas Hellstrom
2006-10-10only allow specific type-3 packets to pass the verifier instead of all for ↵Roland Scheidegger
r100/r200 as others might be unsafe (r300 already does this), and add checking for these we need but aren't safe. Check the RADEON_CP_INDX_BUFFER packet on both r200 and r300 as it isn't safe neither.
2006-10-03Get rid of all ugly PTE hacks.Thomas Hellstrom
2006-10-02Bug 6242: [mach64] Use private DMA buffers, part #4.George Sapountzis
mach64_state.c: convert the DRM_MACH64_BLIT ioctl to submit a pointer to user-space memory rather than a DMA buffer index, similar to DRM_MACH64_VERTEX. This change allows the DDX to map the DMA buffers read-only and eliminate a security problem where a client can alter the contents of the DMA buffer after submission to the DRM. This change also affects the DRI/DRM interface. Performace-wise, it basically affects PCI mode where I get a ~12% speedup for some Mesa demos I tested. This is mainly due to eliminating an ioctl for allocating the DMA buffer. mach64_dma.c: move the responsibility for allocating memory for the DMA ring in PCI mode to the DDX. This change affects the DDX/DRM interface and unifies a couple of PCI/AGP code paths for ring memory in the DRM. Bump the mach64 DRM version major and date.
2006-10-02Bug 6242: [mach64] Use private DMA buffers, part #3.George Sapountzis
Add DRM_PCI_BUFFER_RO flag for mapping PCI DMA buffer read-only. An additional flag is needed, since PCI DMA buffers do not have an associated map.
2006-10-02Bug 6242: [mach64] Use private DMA buffers, part #2.George Sapountzis
Factor out from mach64_dma_dispatch_vertex() the code to reclaim an unsed buffer, in preperation for using it in mach64_dma_dispatch_blit() also.
2006-10-02Bug 6242: [mach64] Use private DMA buffers, part #1.George Sapountzis
Factor out from mach64_freelist_get() the code to reclaim a completed buffer, this is to improve readability for me.
2006-10-02Bug 6209: [mach64] AGP DMA buffers not mapped correctly.George Sapountzis
Map the DMA buffers from the same linear address as the vertex bufs. If dev->agp_buffer_token is not set, mach64 drm maps the DMA buffers from linear address 0x0.
2006-10-02Fix type of second argument to spin_lock_irqsave().Michel Dänzer
(cherry picked from f6238cf6244b32bd84e3d2819963d7f5473867c8 commit)
2006-10-02Fix type of second argument to spin_lock_irqsave().Michel Dänzer
2006-10-02Make the user_token 44-bit for TTMs, and have them occupy a unique file spaceThomas Hellstrom
starting at 0x00100000000. This will hopefully allow us to use unmap_mapping_range(). Note that user-space will need 64-bit file offset support.
2006-10-02Add a buffer object manager for TTM maps.Thomas Hellstrom
2006-10-02Allow for 44 bit user-tokens (or drm_file offsets)Thomas Hellstrom
2006-10-02Add a comment to previos commit.Thomas Hellstrom
2006-10-02Trap and be verbose about a deadlock that occurs with AIGLX and drivers thatThomas Hellstrom
use drm_reclaim_buffers_locked().
2006-10-02drm_rmdraw: Declare id and idx as signed so testing for < 0 works as intended.Felix Kühling
(cherry picked from d58389968124191a546a14b42ef84edc224be23d commit)
2006-10-02Make locked tasklet handling more robust.Michel Dänzer
Initialize the spinlock unconditionally when struct drm_device is filled in, and return early in drm_locked_tasklet() if the driver doesn't support IRQs.
2006-10-02drm_rmdraw: Declare id and idx as signed so testing for < 0 works as intended.Felix Kühling
2006-09-29Bump driver date.Thomas Hellstrom
2006-09-29i915: Only schedule vblank tasklet if there are scheduled swaps pending.Michel Dänzer
This fixes issues on X server startup with versions of xf86-video-intel that enable the IRQ before they have a context ID. (cherry picked from 7af93dd9849442270ec89cb4bbeef5bfd4f9e424 commit)
2006-09-29i915: Only initialize IRQ fields in postinstall, not the PIPE_SET ioctl.Michel Dänzer
Some other minor changes in preparation for actually disabling user interrupts.