From 90bdad6ae7e36a302e0e1172e9d4529e66c7dd90 Mon Sep 17 00:00:00 2001 From: jsg Date: Sat, 19 Nov 2016 05:36:14 +0000 Subject: [PATCH] Import libdrm 2.4.73 --- lib/libdrm/Makefile.sources | 1 + lib/libdrm/amdgpu/Makefile.sources | 1 + lib/libdrm/amdgpu/amdgpu-symbol-check | 1 + lib/libdrm/amdgpu/amdgpu.h | 10 + lib/libdrm/amdgpu/amdgpu_asic_id.h | 165 +++ lib/libdrm/amdgpu/amdgpu_device.c | 15 + lib/libdrm/amdgpu/amdgpu_gpu_info.c | 2 + lib/libdrm/freedreno/freedreno-symbol-check | 1 + lib/libdrm/freedreno/freedreno_drmif.h | 1 + lib/libdrm/freedreno/freedreno_priv.h | 3 +- lib/libdrm/freedreno/freedreno_ringbuffer.c | 11 +- lib/libdrm/freedreno/freedreno_ringbuffer.h | 5 + lib/libdrm/freedreno/kgsl/kgsl_ringbuffer.c | 6 +- lib/libdrm/freedreno/msm/msm_drm.h | 22 +- lib/libdrm/freedreno/msm/msm_ringbuffer.c | 18 +- lib/libdrm/include/drm/README | 157 +++ lib/libdrm/include/drm/drm_mode.h | 49 +- lib/libdrm/intel/Makefile.sources | 3 +- lib/libdrm/intel/intel-symbol-check | 3 + lib/libdrm/intel/uthash.h | 1074 +++++++++++++++++++ lib/libdrm/libsync.h | 148 +++ 21 files changed, 1684 insertions(+), 12 deletions(-) create mode 100644 lib/libdrm/amdgpu/amdgpu_asic_id.h create mode 100644 lib/libdrm/include/drm/README create mode 100644 lib/libdrm/intel/uthash.h create mode 100644 lib/libdrm/libsync.h diff --git a/lib/libdrm/Makefile.sources b/lib/libdrm/Makefile.sources index a57036a68..10aa1d0f4 100644 --- a/lib/libdrm/Makefile.sources +++ b/lib/libdrm/Makefile.sources @@ -13,6 +13,7 @@ LIBDRM_FILES := \ util_math.h LIBDRM_H_FILES := \ + libsync.h \ xf86drm.h \ xf86drmMode.h diff --git a/lib/libdrm/amdgpu/Makefile.sources b/lib/libdrm/amdgpu/Makefile.sources index 0c0b9a93a..487b9e0ad 100644 --- a/lib/libdrm/amdgpu/Makefile.sources +++ b/lib/libdrm/amdgpu/Makefile.sources @@ -1,4 +1,5 @@ LIBDRM_AMDGPU_FILES := \ + amdgpu_asic_id.h \ amdgpu_bo.c \ amdgpu_cs.c \ amdgpu_device.c \ diff --git a/lib/libdrm/amdgpu/amdgpu-symbol-check b/lib/libdrm/amdgpu/amdgpu-symbol-check index 648db9b6c..87f4fd2cc 100755 --- a/lib/libdrm/amdgpu/amdgpu-symbol-check +++ b/lib/libdrm/amdgpu/amdgpu-symbol-check @@ -35,6 +35,7 @@ amdgpu_cs_submit amdgpu_cs_wait_semaphore amdgpu_device_deinitialize amdgpu_device_initialize +amdgpu_get_marketing_name amdgpu_query_buffer_size_alignment amdgpu_query_crtc_from_id amdgpu_query_firmware_version diff --git a/lib/libdrm/amdgpu/amdgpu.h b/lib/libdrm/amdgpu/amdgpu.h index 5d5a2c67f..7b26a04c9 100644 --- a/lib/libdrm/amdgpu/amdgpu.h +++ b/lib/libdrm/amdgpu/amdgpu.h @@ -1245,4 +1245,14 @@ int amdgpu_cs_wait_semaphore(amdgpu_context_handle ctx, */ int amdgpu_cs_destroy_semaphore(amdgpu_semaphore_handle sem); +/** + * Get the ASIC marketing name + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * + * \return the constant string of the marketing name + * "NULL" means the ASIC is not found +*/ +const char *amdgpu_get_marketing_name(amdgpu_device_handle dev); + #endif /* #ifdef _AMDGPU_H_ */ diff --git a/lib/libdrm/amdgpu/amdgpu_asic_id.h b/lib/libdrm/amdgpu/amdgpu_asic_id.h new file mode 100644 index 000000000..3e7d736b6 --- /dev/null +++ b/lib/libdrm/amdgpu/amdgpu_asic_id.h @@ -0,0 +1,165 @@ +/* + * Copyright © 2016 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_ASIC_ID_H__ +#define __AMDGPU_ASIC_ID_H__ + +static struct amdgpu_asic_id_table_t { + uint32_t did; + uint32_t rid; + const char *marketing_name; +} const amdgpu_asic_id_table [] = { + {0x6600, 0x0, "AMD Radeon HD 8600/8700M"}, + {0x6600, 0x81, "AMD Radeon R7 M370"}, + {0x6601, 0x0, "AMD Radeon HD 8500M/8700M"}, + {0x6604, 0x0, "AMD Radeon R7 M265 Series"}, + {0x6604, 0x81, "AMD Radeon R7 M350"}, + {0x6605, 0x0, "AMD Radeon R7 M260 Series"}, + {0x6605, 0x81, "AMD Radeon R7 M340"}, + {0x6606, 0x0, "AMD Radeon HD 8790M"}, + {0x6607, 0x0, "AMD Radeon HD8530M"}, + {0x6608, 0x0, "AMD FirePro W2100"}, + {0x6610, 0x0, "AMD Radeon HD 8600 Series"}, + {0x6610, 0x81, "AMD Radeon R7 350"}, + {0x6610, 0x83, "AMD Radeon R5 340"}, + {0x6611, 0x0, "AMD Radeon HD 8500 Series"}, + {0x6613, 0x0, "AMD Radeon HD 8500 series"}, + {0x6617, 0xC7, "AMD Radeon R7 240 Series"}, + {0x6640, 0x0, "AMD Radeon HD 8950"}, + {0x6640, 0x80, "AMD Radeon R9 M380"}, + {0x6646, 0x0, "AMD Radeon R9 M280X"}, + {0x6646, 0x80, "AMD Radeon R9 M470X"}, + {0x6647, 0x0, "AMD Radeon R9 M270X"}, + {0x6647, 0x80, "AMD Radeon R9 M380"}, + {0x6649, 0x0, "AMD FirePro W5100"}, + {0x6658, 0x0, "AMD Radeon R7 200 Series"}, + {0x665C, 0x0, "AMD Radeon HD 7700 Series"}, + {0x665D, 0x0, "AMD Radeon R7 200 Series"}, + {0x665F, 0x81, "AMD Radeon R7 300 Series"}, + {0x6660, 0x0, "AMD Radeon HD 8600M Series"}, + {0x6660, 0x81, "AMD Radeon R5 M335"}, + {0x6660, 0x83, "AMD Radeon R5 M330"}, + {0x6663, 0x0, "AMD Radeon HD 8500M Series"}, + {0x6663, 0x83, "AMD Radeon R5 M320"}, + {0x6664, 0x0, "AMD Radeon R5 M200 Series"}, + {0x6665, 0x0, "AMD Radeon R5 M200 Series"}, + {0x6665, 0x83, "AMD Radeon R5 M320"}, + {0x6667, 0x0, "AMD Radeon R5 M200 Series"}, + {0x666F, 0x0, "AMD Radeon HD 8500M"}, + {0x6780, 0x0, "ATI FirePro V (FireGL V) Graphics Adapter"}, + {0x678A, 0x0, "ATI FirePro V (FireGL V) Graphics Adapter"}, + {0x6798, 0x0, "AMD Radeon HD 7900 Series"}, + {0x679A, 0x0, "AMD Radeon HD 7900 Series"}, + {0x679B, 0x0, "AMD Radeon HD 7900 Series"}, + {0x679E, 0x0, "AMD Radeon HD 7800 Series"}, + {0x67A0, 0x0, "HAWAII XTGL (67A0)"}, + {0x67A1, 0x0, "HAWAII GL40 (67A1)"}, + {0x67B0, 0x0, "AMD Radeon R9 200 Series"}, + {0x67B0, 0x80, "AMD Radeon R9 390 Series"}, + {0x67B1, 0x0, "AMD Radeon R9 200 Series"}, + {0x67B1, 0x80, "AMD Radeon R9 390 Series"}, + {0x67B9, 0x0, "AMD Radeon R9 200 Series"}, + {0x67DF, 0xC4, "AMD Radeon RX 480 Graphics"}, + {0x67DF, 0xC5, "AMD Radeon RX 470 Graphics"}, + {0x67DF, 0xC7, "AMD Radeon RX 480 Graphics"}, + {0x67DF, 0xCF, "AMD Radeon RX 470 Graphics"}, + {0x67C4, 0x00, "AMD Radeon Pro WX 7100 Graphics"}, + {0x67C7, 0x00, "AMD Radeon Pro WX 5100 Graphics"}, + {0x67C0, 0x00, "AMD Radeon Pro WX 7100 Graphics"}, + {0x67E0, 0x00, "AMD Radeon Pro WX Series Graphics"}, + {0x67E3, 0x00, "AMD Radeon Pro WX 4100 Graphics"}, + {0x67E8, 0x00, "AMD Radeon Pro WX Series Graphics"}, + {0x67E8, 0x01, "AMD Radeon Pro WX Series Graphics"}, + {0x67E8, 0x80, "AMD Radeon E9260 Graphics"}, + {0x67EB, 0x00, "AMD Radeon Pro WX Series Graphics"}, + {0x67EF, 0xC0, "AMD Radeon RX Graphics"}, + {0x67EF, 0xC1, "AMD Radeon RX 460 Graphics"}, + {0x67EF, 0xC5, "AMD Radeon RX 460 Graphics"}, + {0x67EF, 0xC7, "AMD Radeon RX Graphics"}, + {0x67EF, 0xCF, "AMD Radeon RX 460 Graphics"}, + {0x67EF, 0xEF, "AMD Radeon RX Graphics"}, + {0x67FF, 0xC0, "AMD Radeon RX Graphics"}, + {0x67FF, 0xC1, "AMD Radeon RX Graphics"}, + {0x6800, 0x0, "AMD Radeon HD 7970M"}, + {0x6801, 0x0, "AMD Radeon(TM) HD8970M"}, + {0x6808, 0x0, "ATI FirePro V(FireGL V) Graphics Adapter"}, + {0x6809, 0x0, "ATI FirePro V(FireGL V) Graphics Adapter"}, + {0x6810, 0x0, "AMD Radeon(TM) HD 8800 Series"}, + {0x6810, 0x81, "AMD Radeon R7 370 Series"}, + {0x6811, 0x0, "AMD Radeon(TM) HD8800 Series"}, + {0x6811, 0x81, "AMD Radeon R7 300 Series"}, + {0x6818, 0x0, "AMD Radeon HD 7800 Series"}, + {0x6819, 0x0, "AMD Radeon HD 7800 Series"}, + {0x6820, 0x0, "AMD Radeon HD 8800M Series"}, + {0x6820, 0x81, "AMD Radeon R9 M375"}, + {0x6820, 0x83, "AMD Radeon R9 M375X"}, + {0x6821, 0x0, "AMD Radeon HD 8800M Series"}, + {0x6821, 0x87, "AMD Radeon R7 M380"}, + {0x6821, 0x83, "AMD Radeon R9 M370X"}, + {0x6822, 0x0, "AMD Radeon E8860"}, + {0x6823, 0x0, "AMD Radeon HD 8800M Series"}, + {0x6825, 0x0, "AMD Radeon HD 7800M Series"}, + {0x6827, 0x0, "AMD Radeon HD 7800M Series"}, + {0x6828, 0x0, "ATI FirePro V(FireGL V) Graphics Adapter"}, + {0x682B, 0x0, "AMD Radeon HD 8800M Series"}, + {0x682B, 0x87, "AMD Radeon R9 M360"}, + {0x682C, 0x0, "AMD FirePro W4100"}, + {0x682D, 0x0, "AMD Radeon HD 7700M Series"}, + {0x682F, 0x0, "AMD Radeon HD 7700M Series"}, + {0x6835, 0x0, "AMD Radeon R7 Series / HD 9000 Series"}, + {0x6837, 0x0, "AMD Radeon HD7700 Series"}, + {0x683D, 0x0, "AMD Radeon HD 7700 Series"}, + {0x683F, 0x0, "AMD Radeon HD 7700 Series"}, + {0x6900, 0x0, "AMD Radeon R7 M260"}, + {0x6900, 0x81, "AMD Radeon R7 M360"}, + {0x6900, 0x83, "AMD Radeon R7 M340"}, + {0x6901, 0x0, "AMD Radeon R5 M255"}, + {0x6907, 0x0, "AMD Radeon R5 M255"}, + {0x6907, 0x87, "AMD Radeon R5 M315"}, + {0x6920, 0x0, "AMD Radeon R9 M395X"}, + {0x6920, 0x1, "AMD Radeon R9 M390X"}, + {0x6921, 0x0, "AMD Radeon R9 M295X"}, + {0x6929, 0x0, "AMD FirePro S7150"}, + {0x692B, 0x0, "AMD FirePro W7100"}, + {0x6938, 0x0, "AMD Radeon R9 200 Series"}, + {0x6938, 0xF0, "AMD Radeon R9 200 Series"}, + {0x6938, 0xF1, "AMD Radeon R9 380 Series"}, + {0x6939, 0xF0, "AMD Radeon R9 200 Series"}, + {0x6939, 0x0, "AMD Radeon R9 200 Series"}, + {0x6939, 0xF1, "AMD Radeon R9 380 Series"}, + {0x7300, 0xC8, "AMD Radeon R9 Fury Series"}, + {0x7300, 0xCB, "AMD Radeon R9 Fury Series"}, + {0x7300, 0xCA, "AMD Radeon R9 Fury Series"}, + {0x9874, 0xC4, "AMD Radeon R7 Graphics"}, + {0x9874, 0xC5, "AMD Radeon R6 Graphics"}, + {0x9874, 0xC6, "AMD Radeon R6 Graphics"}, + {0x9874, 0xC7, "AMD Radeon R5 Graphics"}, + {0x9874, 0x81, "AMD Radeon R6 Graphics"}, + {0x9874, 0x87, "AMD Radeon R5 Graphics"}, + {0x9874, 0x85, "AMD Radeon R6 Graphics"}, + {0x9874, 0x84, "AMD Radeon R7 Graphics"}, + + {0x0000, 0x0, "\0"}, +}; +#endif diff --git a/lib/libdrm/amdgpu/amdgpu_device.c b/lib/libdrm/amdgpu/amdgpu_device.c index e5a923e67..f4ede0316 100644 --- a/lib/libdrm/amdgpu/amdgpu_device.c +++ b/lib/libdrm/amdgpu/amdgpu_device.c @@ -44,6 +44,7 @@ #include "amdgpu_internal.h" #include "util_hash_table.h" #include "util_math.h" +#include "amdgpu_asic_id.h" #define PTR_TO_UINT(x) ((unsigned)((intptr_t)(x))) #define UINT_TO_PTR(x) ((void *)((intptr_t)(x))) @@ -303,3 +304,17 @@ int amdgpu_device_deinitialize(amdgpu_device_handle dev) amdgpu_device_reference(&dev, NULL); return 0; } + +const char *amdgpu_get_marketing_name(amdgpu_device_handle dev) +{ + const struct amdgpu_asic_id_table_t *t = amdgpu_asic_id_table; + + while (t->did) { + if ((t->did == dev->info.asic_id) && + (t->rid == dev->info.pci_rev_id)) + return t->marketing_name; + t++; + } + + return NULL; +} diff --git a/lib/libdrm/amdgpu/amdgpu_gpu_info.c b/lib/libdrm/amdgpu/amdgpu_gpu_info.c index 0f232977f..66c7e0e1b 100644 --- a/lib/libdrm/amdgpu/amdgpu_gpu_info.c +++ b/lib/libdrm/amdgpu/amdgpu_gpu_info.c @@ -230,6 +230,8 @@ drm_private int amdgpu_query_gpu_info_init(amdgpu_device_handle dev) int amdgpu_query_gpu_info(amdgpu_device_handle dev, struct amdgpu_gpu_info *info) { + if ((dev == NULL) || (info == NULL)) + return -EINVAL; /* Get ASIC info*/ *info = dev->info; diff --git a/lib/libdrm/freedreno/freedreno-symbol-check b/lib/libdrm/freedreno/freedreno-symbol-check index 7b31a3470..ad367fc34 100755 --- a/lib/libdrm/freedreno/freedreno-symbol-check +++ b/lib/libdrm/freedreno/freedreno-symbol-check @@ -49,6 +49,7 @@ fd_ringbuffer_timestamp fd_ringmarker_del fd_ringmarker_dwords fd_ringmarker_flush +fd_ringbuffer_flush2 fd_ringmarker_mark fd_ringmarker_new EOF diff --git a/lib/libdrm/freedreno/freedreno_drmif.h b/lib/libdrm/freedreno/freedreno_drmif.h index 2d913e6ca..7a8073ff7 100644 --- a/lib/libdrm/freedreno/freedreno_drmif.h +++ b/lib/libdrm/freedreno/freedreno_drmif.h @@ -92,6 +92,7 @@ int fd_device_fd(struct fd_device *dev); enum fd_version { FD_VERSION_MADVISE = 1, /* kernel supports madvise */ FD_VERSION_UNLIMITED_CMDS = 1, /* submits w/ >4 cmd buffers (growable ringbuffer) */ + FD_VERSION_FENCE_FD = 2, /* submit command supports in/out fences */ }; enum fd_version fd_device_version(struct fd_device *dev); diff --git a/lib/libdrm/freedreno/freedreno_priv.h b/lib/libdrm/freedreno/freedreno_priv.h index cdfdbe8db..86da83b94 100644 --- a/lib/libdrm/freedreno/freedreno_priv.h +++ b/lib/libdrm/freedreno/freedreno_priv.h @@ -133,7 +133,8 @@ struct fd_ringmarker { struct fd_ringbuffer_funcs { void * (*hostptr)(struct fd_ringbuffer *ring); - int (*flush)(struct fd_ringbuffer *ring, uint32_t *last_start); + int (*flush)(struct fd_ringbuffer *ring, uint32_t *last_start, + int in_fence_fd, int *out_fence_fd); void (*grow)(struct fd_ringbuffer *ring, uint32_t size); void (*reset)(struct fd_ringbuffer *ring); void (*emit_reloc)(struct fd_ringbuffer *ring, diff --git a/lib/libdrm/freedreno/freedreno_ringbuffer.c b/lib/libdrm/freedreno/freedreno_ringbuffer.c index 22dafb394..c132145af 100644 --- a/lib/libdrm/freedreno/freedreno_ringbuffer.c +++ b/lib/libdrm/freedreno/freedreno_ringbuffer.c @@ -80,10 +80,15 @@ void fd_ringbuffer_reset(struct fd_ringbuffer *ring) ring->funcs->reset(ring); } -/* maybe get rid of this and use fd_ringmarker_flush() from DDX too? */ int fd_ringbuffer_flush(struct fd_ringbuffer *ring) { - return ring->funcs->flush(ring, ring->last_start); + return ring->funcs->flush(ring, ring->last_start, -1, NULL); +} + +int fd_ringbuffer_flush2(struct fd_ringbuffer *ring, int in_fence_fd, + int *out_fence_fd) +{ + return ring->funcs->flush(ring, ring->last_start, in_fence_fd, out_fence_fd); } void fd_ringbuffer_grow(struct fd_ringbuffer *ring, uint32_t ndwords) @@ -177,5 +182,5 @@ uint32_t fd_ringmarker_dwords(struct fd_ringmarker *start, int fd_ringmarker_flush(struct fd_ringmarker *marker) { struct fd_ringbuffer *ring = marker->ring; - return ring->funcs->flush(ring, marker->cur); + return ring->funcs->flush(ring, marker->cur, -1, NULL); } diff --git a/lib/libdrm/freedreno/freedreno_ringbuffer.h b/lib/libdrm/freedreno/freedreno_ringbuffer.h index 8899b5de4..108d5a6de 100644 --- a/lib/libdrm/freedreno/freedreno_ringbuffer.h +++ b/lib/libdrm/freedreno/freedreno_ringbuffer.h @@ -56,6 +56,11 @@ void fd_ringbuffer_set_parent(struct fd_ringbuffer *ring, struct fd_ringbuffer *parent); void fd_ringbuffer_reset(struct fd_ringbuffer *ring); int fd_ringbuffer_flush(struct fd_ringbuffer *ring); +/* in_fence_fd: -1 for no in-fence, else fence fd + * out_fence_fd: NULL for no output-fence requested, else ptr to return out-fence + */ +int fd_ringbuffer_flush2(struct fd_ringbuffer *ring, int in_fence_fd, + int *out_fence_fd); void fd_ringbuffer_grow(struct fd_ringbuffer *ring, uint32_t ndwords); uint32_t fd_ringbuffer_timestamp(struct fd_ringbuffer *ring); diff --git a/lib/libdrm/freedreno/kgsl/kgsl_ringbuffer.c b/lib/libdrm/freedreno/kgsl/kgsl_ringbuffer.c index 7b3298abd..e4696b1bf 100644 --- a/lib/libdrm/freedreno/kgsl/kgsl_ringbuffer.c +++ b/lib/libdrm/freedreno/kgsl/kgsl_ringbuffer.c @@ -113,7 +113,8 @@ static void * kgsl_ringbuffer_hostptr(struct fd_ringbuffer *ring) return kgsl_ring->bo->hostptr; } -static int kgsl_ringbuffer_flush(struct fd_ringbuffer *ring, uint32_t *last_start) +static int kgsl_ringbuffer_flush(struct fd_ringbuffer *ring, uint32_t *last_start, + int in_fence_fd, int *out_fence_fd) { struct kgsl_ringbuffer *kgsl_ring = to_kgsl_ringbuffer(ring); struct kgsl_pipe *kgsl_pipe = to_kgsl_pipe(ring->pipe); @@ -131,6 +132,9 @@ static int kgsl_ringbuffer_flush(struct fd_ringbuffer *ring, uint32_t *last_star }; int ret; + assert(in_fence_fd == -1); + assert(out_fence_fd == NULL); + kgsl_pipe_pre_submit(kgsl_pipe); /* z180_cmdstream_issueibcmds() is made of fail: */ diff --git a/lib/libdrm/freedreno/msm/msm_drm.h b/lib/libdrm/freedreno/msm/msm_drm.h index cbf75c3d5..ed4c8d475 100644 --- a/lib/libdrm/freedreno/msm/msm_drm.h +++ b/lib/libdrm/freedreno/msm/msm_drm.h @@ -50,6 +50,15 @@ extern "C" { #define MSM_PIPE_2D1 0x02 #define MSM_PIPE_3D0 0x10 +/* The pipe-id just uses the lower bits, so can be OR'd with flags in + * the upper 16 bits (which could be extended further, if needed, maybe + * we extend/overload the pipe-id some day to deal with multiple rings, + * but even then I don't think we need the full lower 16 bits). + */ +#define MSM_PIPE_ID_MASK 0xffff +#define MSM_PIPE_ID(x) ((x) & MSM_PIPE_ID_MASK) +#define MSM_PIPE_FLAGS(x) ((x) & ~MSM_PIPE_ID_MASK) + /* timeouts are specified in clock-monotonic absolute times (to simplify * restarting interrupted ioctls). The following struct is logically the * same as 'struct timespec' but 32/64b ABI safe. @@ -183,17 +192,28 @@ struct drm_msm_gem_submit_bo { __u64 presumed; /* in/out, presumed buffer address */ }; +/* Valid submit ioctl flags: */ +#define MSM_SUBMIT_NO_IMPLICIT 0x80000000 /* disable implicit sync */ +#define MSM_SUBMIT_FENCE_FD_IN 0x40000000 /* enable input fence_fd */ +#define MSM_SUBMIT_FENCE_FD_OUT 0x20000000 /* enable output fence_fd */ +#define MSM_SUBMIT_FLAGS ( \ + MSM_SUBMIT_NO_IMPLICIT | \ + MSM_SUBMIT_FENCE_FD_IN | \ + MSM_SUBMIT_FENCE_FD_OUT | \ + 0) + /* Each cmdstream submit consists of a table of buffers involved, and * one or more cmdstream buffers. This allows for conditional execution * (context-restore), and IB buffers needed for per tile/bin draw cmds. */ struct drm_msm_gem_submit { - __u32 pipe; /* in, MSM_PIPE_x */ + __u32 flags; /* MSM_PIPE_x | MSM_SUBMIT_x */ __u32 fence; /* out */ __u32 nr_bos; /* in, number of submit_bo's */ __u32 nr_cmds; /* in, number of submit_cmd's */ __u64 __user bos; /* in, ptr to array of submit_bo's */ __u64 __user cmds; /* in, ptr to array of submit_cmd's */ + __s32 fence_fd; /* in/out fence fd (see MSM_SUBMIT_FENCE_FD_IN/OUT) */ }; /* The normal way to synchronize with the GPU is just to CPU_PREP on diff --git a/lib/libdrm/freedreno/msm/msm_ringbuffer.c b/lib/libdrm/freedreno/msm/msm_ringbuffer.c index a78806c88..5117df1a9 100644 --- a/lib/libdrm/freedreno/msm/msm_ringbuffer.c +++ b/lib/libdrm/freedreno/msm/msm_ringbuffer.c @@ -395,15 +395,25 @@ static void dump_submit(struct msm_ringbuffer *msm_ring) } } -static int msm_ringbuffer_flush(struct fd_ringbuffer *ring, uint32_t *last_start) +static int msm_ringbuffer_flush(struct fd_ringbuffer *ring, uint32_t *last_start, + int in_fence_fd, int *out_fence_fd) { struct msm_ringbuffer *msm_ring = to_msm_ringbuffer(ring); struct drm_msm_gem_submit req = { - .pipe = to_msm_pipe(ring->pipe)->pipe, + .flags = to_msm_pipe(ring->pipe)->pipe, }; uint32_t i; int ret; + if (in_fence_fd != -1) { + req.flags |= MSM_SUBMIT_FENCE_FD_IN | MSM_SUBMIT_NO_IMPLICIT; + req.fence_fd = in_fence_fd; + } + + if (out_fence_fd) { + req.flags |= MSM_SUBMIT_FENCE_FD_OUT; + } + finalize_current_cmd(ring, last_start); /* needs to be after get_cmd() as that could create bos/cmds table: */ @@ -435,6 +445,10 @@ static int msm_ringbuffer_flush(struct fd_ringbuffer *ring, uint32_t *last_start struct msm_cmd *msm_cmd = msm_ring->cmds[i]; msm_cmd->ring->last_timestamp = req.fence; } + + if (out_fence_fd) { + *out_fence_fd = req.fence_fd; + } } flush_reset(ring); diff --git a/lib/libdrm/include/drm/README b/lib/libdrm/include/drm/README new file mode 100644 index 000000000..c3292f312 --- /dev/null +++ b/lib/libdrm/include/drm/README @@ -0,0 +1,157 @@ +What are these headers ? +------------------------ +This is the canonical source of drm headers that user space should use for +communicating with the kernel DRM subsystem. + +They flow from the kernel, thus any changes must be merged there first. +Do _not_ attempt to "fix" these by deviating from the kernel ones ! + + +Non-linux platforms - changes/patches +------------------------------------- +If your platform has local changes, please send them upstream for inclusion. +Even if your patches don't get accepted in their current form, devs will +give you feedback on how to address things properly. + +git send-email --subject-prefix="PATCH libdrm" your patches to dri-devel +mailing list. + +Before doing so, please consider the following: + - Have the [libdrm vs kernel] headers on your platform deviated ? +Consider unifying them first. + + - Have you introduced additional ABI that's not available in Linux ? +Propose it for [Linux kernel] upstream inclusion. +If that doesn't work out (hopefully it never does), move it to another header +and/or keep the change(s) local ? + + - Are your changes DRI1/UMS specific ? +There is virtually no interest/power in keeping those legacy interfaces. They +are around due to the kernel "thou shalt not break existing user space" rule. + +Consider porting the driver to DRI2/KMS - all (almost?) sensible hardware is +capable of supporting those. + + +Which headers go where ? +------------------------ +A snipped from the, now removed, Makefile.am used to state: + + XXX airlied says, nothing besides *_drm.h and drm*.h should be necessary. + however, r300 and via need their reg headers installed in order to build. + better solutions are welcome. + +Obviously the r300 and via headers are no longer around ;-) + +Reason behind is that the drm headers can be used as a basic communications +channel with the respective kernel modules. If more advanced functionality is +required one can pull the specific libdrm_$driver which is free to pull +additional files from the kernel. + +For example: nouveau has nouveau/nvif/*.h while vc4 has vc4/*.h + +If your driver is still in prototyping/staging state, consider moving the +$driver_drm.h into $driver and _not_ installing it. An header providing opaque +definitions and access [via $driver_drmif.h or similar] would be better fit. + + +When and which headers to update +-------------------------------- +Ideally all files will be synced (updated) with the latest released kernel on +each libdrm release. Sadly that's not yet possible since quite a few headers +differ significantly - see Outdated or Broken Headers section below. + +That said, it's up-to the individual developers to sync with newer version +(from drm-next) as they see fit. + + +When and how to update these files +---------------------------------- +In order to update the files do the following: + - Switch to a Linux kernel tree/branch which is not rebased. +For example: airlied/drm-next + - Install the headers via `make headers_install' to a separate location. + - Copy the drm header[s] + git add + git commit. + - Note: Your commit message must include: + a) Brief summary on the delta. If there's any change that looks like an +API/ABI break one _must_ explicitly state why it's safe to do so. + b) "Generated using make headers_install." + c) "Generated from $tree/branch commit $sha" + + +Outdated or Broken Headers +-------------------------- +This section contains a list of headers and the respective "issues" they might +have relative to their kernel equivalent. + +Nearly all headers: + - Missing extern C notation. +Status: Trivial. + +Most UMS headers: + - Not using fixed size interers - compat ioctls are broken. +Status: ? +Promote to fixed size ints, which match the current (32bit) ones. + + +amdgpu_drm.h + - Using the stdint.h uint*_t over the respective __u* ones +Status: Trivial. + +drm_mode.h + - Missing DPI encode/connector pair. +Status: Trivial. + +i915_drm.h + - Missing PARAMS - HAS_POOLED_EU, MIN_EU_IN_POOL CONTEXT_PARAM_NO_ERROR_CAPTURE +Status: Trivial. + +mga_drm.h + - Typo fix, use struct over typedef. +Status: Trivial. + +nouveau_drm.h + - Missing macros NOUVEAU_GETPARAM*, NOUVEAU_DRM_HEADER_PATCHLEVEL, structs, +enums, using stdint.h over the __u* types. +Status: ? + +qxl_drm.h + - Using the stdint.h uint*_t over the respective __u* ones +Status: Trivial. + +r128_drm.h + - Broken compat ioctls. + +radeon_drm.h + - Missing RADEON_TILING_R600_NO_SCANOUT, CIK_TILE_MODE_*, broken UMS ioctls, +using stdint types. + - Both kernel and libdrm: missing padding - +drm_radeon_gem_{create,{g,s}et_tiling,set_domain} others ? +Status: ? + +savage_drm.h + - Renamed ioctls - DRM_IOCTL_SAVAGE_{,BCI}_EVENT_EMIT, compat ioctls are broken. +Status: ? + +sis_drm.h + - Borken ioctls + libdrm uses int vs kernel long +Status: ? + +via_drm.h + - Borken ioctls - libdrm int vs kernel long +Status: ? + + +omap_drm.h (living in $TOP/omap) + - License mismatch, missing DRM_IOCTL_OMAP_GEM_NEW and related struct +Status: ? + +msm_drm.h (located in $TOP/freedreno/msm/) + - License mismatch, missing MSM_PIPE_*, MSM_SUBMIT_*. Renamed +drm_msm_gem_submit::flags, missing drm_msm_gem_submit::fence_fd. +Status: ? + +exynos_drm.h (living in $TOP/exynos) + - License mismatch, now using fixed size ints (but not everywhere). Lots of +new stuff. +Status: ? diff --git a/lib/libdrm/include/drm/drm_mode.h b/lib/libdrm/include/drm/drm_mode.h index 7a7856e02..df0e3504c 100644 --- a/lib/libdrm/include/drm/drm_mode.h +++ b/lib/libdrm/include/drm/drm_mode.h @@ -29,6 +29,10 @@ #include "drm.h" +#if defined(__cplusplus) +extern "C" { +#endif + #define DRM_DISPLAY_INFO_LEN 32 #define DRM_CONNECTOR_NAME_LEN 32 #define DRM_DISPLAY_MODE_LEN 32 @@ -202,6 +206,7 @@ struct drm_mode_get_plane_res { #define DRM_MODE_ENCODER_VIRTUAL 5 #define DRM_MODE_ENCODER_DSI 6 #define DRM_MODE_ENCODER_DPMST 7 +#define DRM_MODE_ENCODER_DPI 8 struct drm_mode_get_encoder { __u32 encoder_id; @@ -241,6 +246,7 @@ struct drm_mode_get_encoder { #define DRM_MODE_CONNECTOR_eDP 14 #define DRM_MODE_CONNECTOR_VIRTUAL 15 #define DRM_MODE_CONNECTOR_DSI 16 +#define DRM_MODE_CONNECTOR_DPI 17 struct drm_mode_get_connector { @@ -514,7 +520,13 @@ struct drm_color_lut { #define DRM_MODE_PAGE_FLIP_EVENT 0x01 #define DRM_MODE_PAGE_FLIP_ASYNC 0x02 -#define DRM_MODE_PAGE_FLIP_FLAGS (DRM_MODE_PAGE_FLIP_EVENT|DRM_MODE_PAGE_FLIP_ASYNC) +#define DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE 0x4 +#define DRM_MODE_PAGE_FLIP_TARGET_RELATIVE 0x8 +#define DRM_MODE_PAGE_FLIP_TARGET (DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE | \ + DRM_MODE_PAGE_FLIP_TARGET_RELATIVE) +#define DRM_MODE_PAGE_FLIP_FLAGS (DRM_MODE_PAGE_FLIP_EVENT | \ + DRM_MODE_PAGE_FLIP_ASYNC | \ + DRM_MODE_PAGE_FLIP_TARGET) /* * Request a page flip on the specified crtc. @@ -537,8 +549,7 @@ struct drm_color_lut { * 'as soon as possible', meaning that it not delay waiting for vblank. * This may cause tearing on the screen. * - * The reserved field must be zero until we figure out something - * clever to use it for. + * The reserved field must be zero. */ struct drm_mode_crtc_page_flip { @@ -549,6 +560,34 @@ struct drm_mode_crtc_page_flip { __u64 user_data; }; +/* + * Request a page flip on the specified crtc. + * + * Same as struct drm_mode_crtc_page_flip, but supports new flags and + * re-purposes the reserved field: + * + * The sequence field must be zero unless either of the + * DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE/RELATIVE flags is specified. When + * the ABSOLUTE flag is specified, the sequence field denotes the absolute + * vblank sequence when the flip should take effect. When the RELATIVE + * flag is specified, the sequence field denotes the relative (to the + * current one when the ioctl is called) vblank sequence when the flip + * should take effect. NOTE: DRM_IOCTL_WAIT_VBLANK must still be used to + * make sure the vblank sequence before the target one has passed before + * calling this ioctl. The purpose of the + * DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE/RELATIVE flags is merely to clarify + * the target for when code dealing with a page flip runs during a + * vertical blank period. + */ + +struct drm_mode_crtc_page_flip_target { + __u32 crtc_id; + __u32 fb_id; + __u32 flags; + __u32 sequence; + __u64 user_data; +}; + /* create a dumb scanout buffer */ struct drm_mode_create_dumb { __u32 height; @@ -621,4 +660,8 @@ struct drm_mode_destroy_blob { __u32 blob_id; }; +#if defined(__cplusplus) +} +#endif + #endif diff --git a/lib/libdrm/intel/Makefile.sources b/lib/libdrm/intel/Makefile.sources index 7b2272c75..6947ab742 100644 --- a/lib/libdrm/intel/Makefile.sources +++ b/lib/libdrm/intel/Makefile.sources @@ -6,7 +6,8 @@ LIBDRM_INTEL_FILES := \ intel_decode.c \ intel_chipset.h \ mm.c \ - mm.h + mm.h \ + uthash.h LIBDRM_INTEL_H_FILES := \ intel_bufmgr.h \ diff --git a/lib/libdrm/intel/intel-symbol-check b/lib/libdrm/intel/intel-symbol-check index fd63b3737..4462533c2 100755 --- a/lib/libdrm/intel/intel-symbol-check +++ b/lib/libdrm/intel/intel-symbol-check @@ -70,6 +70,9 @@ drm_intel_gem_bo_aub_dump_bmp drm_intel_gem_bo_clear_relocs drm_intel_gem_bo_context_exec drm_intel_gem_bo_get_reloc_count +drm_intel_gem_bo_map__cpu +drm_intel_gem_bo_map__gtt +drm_intel_gem_bo_map__wc drm_intel_gem_bo_map_gtt drm_intel_gem_bo_map_unsynchronized drm_intel_gem_bo_start_gtt_access diff --git a/lib/libdrm/intel/uthash.h b/lib/libdrm/intel/uthash.h new file mode 100644 index 000000000..45d1f9fc1 --- /dev/null +++ b/lib/libdrm/intel/uthash.h @@ -0,0 +1,1074 @@ +/* +Copyright (c) 2003-2016, Troy D. Hanson http://troydhanson.github.com/uthash/ +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS +IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED +TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A +PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef UTHASH_H +#define UTHASH_H + +#define UTHASH_VERSION 2.0.1 + +#include /* memcmp,strlen */ +#include /* ptrdiff_t */ +#include /* exit() */ + +/* These macros use decltype or the earlier __typeof GNU extension. + As decltype is only available in newer compilers (VS2010 or gcc 4.3+ + when compiling c++ source) this code uses whatever method is needed + or, for VS2008 where neither is available, uses casting workarounds. */ +#if defined(_MSC_VER) /* MS compiler */ +#if _MSC_VER >= 1600 && defined(__cplusplus) /* VS2010 or newer in C++ mode */ +#define DECLTYPE(x) (decltype(x)) +#else /* VS2008 or older (or VS2010 in C mode) */ +#define NO_DECLTYPE +#define DECLTYPE(x) +#endif +#elif defined(__BORLANDC__) || defined(__LCC__) || defined(__WATCOMC__) +#define NO_DECLTYPE +#define DECLTYPE(x) +#else /* GNU, Sun and other compilers */ +#define DECLTYPE(x) (__typeof(x)) +#endif + +#ifdef NO_DECLTYPE +#define DECLTYPE_ASSIGN(dst,src) \ +do { \ + char **_da_dst = (char**)(&(dst)); \ + *_da_dst = (char*)(src); \ +} while (0) +#else +#define DECLTYPE_ASSIGN(dst,src) \ +do { \ + (dst) = DECLTYPE(dst)(src); \ +} while (0) +#endif + +/* a number of the hash function use uint32_t which isn't defined on Pre VS2010 */ +#if defined(_WIN32) +#if defined(_MSC_VER) && _MSC_VER >= 1600 +#include +#elif defined(__WATCOMC__) || defined(__MINGW32__) || defined(__CYGWIN__) +#include +#else +typedef unsigned int uint32_t; +typedef unsigned char uint8_t; +#endif +#elif defined(__GNUC__) && !defined(__VXWORKS__) +#include +#else +typedef unsigned int uint32_t; +typedef unsigned char uint8_t; +#endif + +#ifndef uthash_fatal +#define uthash_fatal(msg) exit(-1) /* fatal error (out of memory,etc) */ +#endif +#ifndef uthash_malloc +#define uthash_malloc(sz) malloc(sz) /* malloc fcn */ +#endif +#ifndef uthash_free +#define uthash_free(ptr,sz) free(ptr) /* free fcn */ +#endif +#ifndef uthash_strlen +#define uthash_strlen(s) strlen(s) +#endif +#ifndef uthash_memcmp +#define uthash_memcmp(a,b,n) memcmp(a,b,n) +#endif + +#ifndef uthash_noexpand_fyi +#define uthash_noexpand_fyi(tbl) /* can be defined to log noexpand */ +#endif +#ifndef uthash_expand_fyi +#define uthash_expand_fyi(tbl) /* can be defined to log expands */ +#endif + +/* initial number of buckets */ +#define HASH_INITIAL_NUM_BUCKETS 32U /* initial number of buckets */ +#define HASH_INITIAL_NUM_BUCKETS_LOG2 5U /* lg2 of initial number of buckets */ +#define HASH_BKT_CAPACITY_THRESH 10U /* expand when bucket count reaches */ + +/* calculate the element whose hash handle address is hhp */ +#define ELMT_FROM_HH(tbl,hhp) ((void*)(((char*)(hhp)) - ((tbl)->hho))) +/* calculate the hash handle from element address elp */ +#define HH_FROM_ELMT(tbl,elp) ((UT_hash_handle *)(((char*)(elp)) + ((tbl)->hho))) + +#define HASH_VALUE(keyptr,keylen,hashv) \ +do { \ + HASH_FCN(keyptr, keylen, hashv); \ +} while (0) + +#define HASH_FIND_BYHASHVALUE(hh,head,keyptr,keylen,hashval,out) \ +do { \ + (out) = NULL; \ + if (head) { \ + unsigned _hf_bkt; \ + HASH_TO_BKT(hashval, (head)->hh.tbl->num_buckets, _hf_bkt); \ + if (HASH_BLOOM_TEST((head)->hh.tbl, hashval) != 0) { \ + HASH_FIND_IN_BKT((head)->hh.tbl, hh, (head)->hh.tbl->buckets[ _hf_bkt ], keyptr, keylen, hashval, out); \ + } \ + } \ +} while (0) + +#define HASH_FIND(hh,head,keyptr,keylen,out) \ +do { \ + unsigned _hf_hashv; \ + HASH_VALUE(keyptr, keylen, _hf_hashv); \ + HASH_FIND_BYHASHVALUE(hh, head, keyptr, keylen, _hf_hashv, out); \ +} while (0) + +#ifdef HASH_BLOOM +#define HASH_BLOOM_BITLEN (1UL << HASH_BLOOM) +#define HASH_BLOOM_BYTELEN (HASH_BLOOM_BITLEN/8UL) + (((HASH_BLOOM_BITLEN%8UL)!=0UL) ? 1UL : 0UL) +#define HASH_BLOOM_MAKE(tbl) \ +do { \ + (tbl)->bloom_nbits = HASH_BLOOM; \ + (tbl)->bloom_bv = (uint8_t*)uthash_malloc(HASH_BLOOM_BYTELEN); \ + if (!((tbl)->bloom_bv)) { uthash_fatal( "out of memory"); } \ + memset((tbl)->bloom_bv, 0, HASH_BLOOM_BYTELEN); \ + (tbl)->bloom_sig = HASH_BLOOM_SIGNATURE; \ +} while (0) + +#define HASH_BLOOM_FREE(tbl) \ +do { \ + uthash_free((tbl)->bloom_bv, HASH_BLOOM_BYTELEN); \ +} while (0) + +#define HASH_BLOOM_BITSET(bv,idx) (bv[(idx)/8U] |= (1U << ((idx)%8U))) +#define HASH_BLOOM_BITTEST(bv,idx) (bv[(idx)/8U] & (1U << ((idx)%8U))) + +#define HASH_BLOOM_ADD(tbl,hashv) \ + HASH_BLOOM_BITSET((tbl)->bloom_bv, (hashv & (uint32_t)((1ULL << (tbl)->bloom_nbits) - 1U))) + +#define HASH_BLOOM_TEST(tbl,hashv) \ + HASH_BLOOM_BITTEST((tbl)->bloom_bv, (hashv & (uint32_t)((1ULL << (tbl)->bloom_nbits) - 1U))) + +#else +#define HASH_BLOOM_MAKE(tbl) +#define HASH_BLOOM_FREE(tbl) +#define HASH_BLOOM_ADD(tbl,hashv) +#define HASH_BLOOM_TEST(tbl,hashv) (1) +#define HASH_BLOOM_BYTELEN 0U +#endif + +#define HASH_MAKE_TABLE(hh,head) \ +do { \ + (head)->hh.tbl = (UT_hash_table*)uthash_malloc( \ + sizeof(UT_hash_table)); \ + if (!((head)->hh.tbl)) { uthash_fatal( "out of memory"); } \ + memset((head)->hh.tbl, 0, sizeof(UT_hash_table)); \ + (head)->hh.tbl->tail = &((head)->hh); \ + (head)->hh.tbl->num_buckets = HASH_INITIAL_NUM_BUCKETS; \ + (head)->hh.tbl->log2_num_buckets = HASH_INITIAL_NUM_BUCKETS_LOG2; \ + (head)->hh.tbl->hho = (char*)(&(head)->hh) - (char*)(head); \ + (head)->hh.tbl->buckets = (UT_hash_bucket*)uthash_malloc( \ + HASH_INITIAL_NUM_BUCKETS*sizeof(struct UT_hash_bucket)); \ + if (! (head)->hh.tbl->buckets) { uthash_fatal( "out of memory"); } \ + memset((head)->hh.tbl->buckets, 0, \ + HASH_INITIAL_NUM_BUCKETS*sizeof(struct UT_hash_bucket)); \ + HASH_BLOOM_MAKE((head)->hh.tbl); \ + (head)->hh.tbl->signature = HASH_SIGNATURE; \ +} while (0) + +#define HASH_REPLACE_BYHASHVALUE_INORDER(hh,head,fieldname,keylen_in,hashval,add,replaced,cmpfcn) \ +do { \ + (replaced) = NULL; \ + HASH_FIND_BYHASHVALUE(hh, head, &((add)->fieldname), keylen_in, hashval, replaced); \ + if (replaced) { \ + HASH_DELETE(hh, head, replaced); \ + } \ + HASH_ADD_KEYPTR_BYHASHVALUE_INORDER(hh, head, &((add)->fieldname), keylen_in, hashval, add, cmpfcn); \ +} while (0) + +#define HASH_REPLACE_BYHASHVALUE(hh,head,fieldname,keylen_in,hashval,add,replaced) \ +do { \ + (replaced) = NULL; \ + HASH_FIND_BYHASHVALUE(hh, head, &((add)->fieldname), keylen_in, hashval, replaced); \ + if (replaced) { \ + HASH_DELETE(hh, head, replaced); \ + } \ + HASH_ADD_KEYPTR_BYHASHVALUE(hh, head, &((add)->fieldname), keylen_in, hashval, add); \ +} while (0) + +#define HASH_REPLACE(hh,head,fieldname,keylen_in,add,replaced) \ +do { \ + unsigned _hr_hashv; \ + HASH_VALUE(&((add)->fieldname), keylen_in, _hr_hashv); \ + HASH_REPLACE_BYHASHVALUE(hh, head, fieldname, keylen_in, _hr_hashv, add, replaced); \ +} while (0) + +#define HASH_REPLACE_INORDER(hh,head,fieldname,keylen_in,add,replaced,cmpfcn) \ +do { \ + unsigned _hr_hashv; \ + HASH_VALUE(&((add)->fieldname), keylen_in, _hr_hashv); \ + HASH_REPLACE_BYHASHVALUE_INORDER(hh, head, fieldname, keylen_in, _hr_hashv, add, replaced, cmpfcn); \ +} while (0) + +#define HASH_APPEND_LIST(hh, head, add) \ +do { \ + (add)->hh.next = NULL; \ + (add)->hh.prev = ELMT_FROM_HH((head)->hh.tbl, (head)->hh.tbl->tail); \ + (head)->hh.tbl->tail->next = (add); \ + (head)->hh.tbl->tail = &((add)->hh); \ +} while (0) + +#define HASH_ADD_KEYPTR_BYHASHVALUE_INORDER(hh,head,keyptr,keylen_in,hashval,add,cmpfcn) \ +do { \ + unsigned _ha_bkt; \ + (add)->hh.hashv = (hashval); \ + (add)->hh.key = (char*) (keyptr); \ + (add)->hh.keylen = (unsigned) (keylen_in); \ + if (!(head)) { \ + (add)->hh.next = NULL; \ + (add)->hh.prev = NULL; \ + (head) = (add); \ + HASH_MAKE_TABLE(hh, head); \ + } else { \ + struct UT_hash_handle *_hs_iter = &(head)->hh; \ + (add)->hh.tbl = (head)->hh.tbl; \ + do { \ + if (cmpfcn(DECLTYPE(head) ELMT_FROM_HH((head)->hh.tbl, _hs_iter), add) > 0) \ + break; \ + } while ((_hs_iter = _hs_iter->next)); \ + if (_hs_iter) { \ + (add)->hh.next = _hs_iter; \ + if (((add)->hh.prev = _hs_iter->prev)) { \ + HH_FROM_ELMT((head)->hh.tbl, _hs_iter->prev)->next = (add); \ + } else { \ + (head) = (add); \ + } \ + _hs_iter->prev = (add); \ + } else { \ + HASH_APPEND_LIST(hh, head, add); \ + } \ + } \ + (head)->hh.tbl->num_items++; \ + HASH_TO_BKT(hashval, (head)->hh.tbl->num_buckets, _ha_bkt); \ + HASH_ADD_TO_BKT((head)->hh.tbl->buckets[_ha_bkt], &(add)->hh); \ + HASH_BLOOM_ADD((head)->hh.tbl, hashval); \ + HASH_EMIT_KEY(hh, head, keyptr, keylen_in); \ + HASH_FSCK(hh, head); \ +} while (0) + +#define HASH_ADD_KEYPTR_INORDER(hh,head,keyptr,keylen_in,add,cmpfcn) \ +do { \ + unsigned _hs_hashv; \ + HASH_VALUE(keyptr, keylen_in, _hs_hashv); \ + HASH_ADD_KEYPTR_BYHASHVALUE_INORDER(hh, head, keyptr, keylen_in, _hs_hashv, add, cmpfcn); \ +} while (0) + +#define HASH_ADD_BYHASHVALUE_INORDER(hh,head,fieldname,keylen_in,hashval,add,cmpfcn) \ + HASH_ADD_KEYPTR_BYHASHVALUE_INORDER(hh, head, &((add)->fieldname), keylen_in, hashval, add, cmpfcn) + +#define HASH_ADD_INORDER(hh,head,fieldname,keylen_in,add,cmpfcn) \ + HASH_ADD_KEYPTR_INORDER(hh, head, &((add)->fieldname), keylen_in, add, cmpfcn) + +#define HASH_ADD_KEYPTR_BYHASHVALUE(hh,head,keyptr,keylen_in,hashval,add) \ +do { \ + unsigned _ha_bkt; \ + (add)->hh.hashv = (hashval); \ + (add)->hh.key = (char*) (keyptr); \ + (add)->hh.keylen = (unsigned) (keylen_in); \ + if (!(head)) { \ + (add)->hh.next = NULL; \ + (add)->hh.prev = NULL; \ + (head) = (add); \ + HASH_MAKE_TABLE(hh, head); \ + } else { \ + (add)->hh.tbl = (head)->hh.tbl; \ + HASH_APPEND_LIST(hh, head, add); \ + } \ + (head)->hh.tbl->num_items++; \ + HASH_TO_BKT(hashval, (head)->hh.tbl->num_buckets, _ha_bkt); \ + HASH_ADD_TO_BKT((head)->hh.tbl->buckets[_ha_bkt], &(add)->hh); \ + HASH_BLOOM_ADD((head)->hh.tbl, hashval); \ + HASH_EMIT_KEY(hh, head, keyptr, keylen_in); \ + HASH_FSCK(hh, head); \ +} while (0) + +#define HASH_ADD_KEYPTR(hh,head,keyptr,keylen_in,add) \ +do { \ + unsigned _ha_hashv; \ + HASH_VALUE(keyptr, keylen_in, _ha_hashv); \ + HASH_ADD_KEYPTR_BYHASHVALUE(hh, head, keyptr, keylen_in, _ha_hashv, add); \ +} while (0) + +#define HASH_ADD_BYHASHVALUE(hh,head,fieldname,keylen_in,hashval,add) \ + HASH_ADD_KEYPTR_BYHASHVALUE(hh, head, &((add)->fieldname), keylen_in, hashval, add) + +#define HASH_ADD(hh,head,fieldname,keylen_in,add) \ + HASH_ADD_KEYPTR(hh, head, &((add)->fieldname), keylen_in, add) + +#define HASH_TO_BKT(hashv,num_bkts,bkt) \ +do { \ + bkt = ((hashv) & ((num_bkts) - 1U)); \ +} while (0) + +/* delete "delptr" from the hash table. + * "the usual" patch-up process for the app-order doubly-linked-list. + * The use of _hd_hh_del below deserves special explanation. + * These used to be expressed using (delptr) but that led to a bug + * if someone used the same symbol for the head and deletee, like + * HASH_DELETE(hh,users,users); + * We want that to work, but by changing the head (users) below + * we were forfeiting our ability to further refer to the deletee (users) + * in the patch-up process. Solution: use scratch space to + * copy the deletee pointer, then the latter references are via that + * scratch pointer rather than through the repointed (users) symbol. + */ +#define HASH_DELETE(hh,head,delptr) \ +do { \ + struct UT_hash_handle *_hd_hh_del; \ + if ( ((delptr)->hh.prev == NULL) && ((delptr)->hh.next == NULL) ) { \ + uthash_free((head)->hh.tbl->buckets, \ + (head)->hh.tbl->num_buckets*sizeof(struct UT_hash_bucket) ); \ + HASH_BLOOM_FREE((head)->hh.tbl); \ + uthash_free((head)->hh.tbl, sizeof(UT_hash_table)); \ + head = NULL; \ + } else { \ + unsigned _hd_bkt; \ + _hd_hh_del = &((delptr)->hh); \ + if ((delptr) == ELMT_FROM_HH((head)->hh.tbl,(head)->hh.tbl->tail)) { \ + (head)->hh.tbl->tail = \ + (UT_hash_handle*)((ptrdiff_t)((delptr)->hh.prev) + \ + (head)->hh.tbl->hho); \ + } \ + if ((delptr)->hh.prev != NULL) { \ + ((UT_hash_handle*)((ptrdiff_t)((delptr)->hh.prev) + \ + (head)->hh.tbl->hho))->next = (delptr)->hh.next; \ + } else { \ + DECLTYPE_ASSIGN(head,(delptr)->hh.next); \ + } \ + if (_hd_hh_del->next != NULL) { \ + ((UT_hash_handle*)((ptrdiff_t)_hd_hh_del->next + \ + (head)->hh.tbl->hho))->prev = \ + _hd_hh_del->prev; \ + } \ + HASH_TO_BKT( _hd_hh_del->hashv, (head)->hh.tbl->num_buckets, _hd_bkt); \ + HASH_DEL_IN_BKT(hh,(head)->hh.tbl->buckets[_hd_bkt], _hd_hh_del); \ + (head)->hh.tbl->num_items--; \ + } \ + HASH_FSCK(hh,head); \ +} while (0) + + +/* convenience forms of HASH_FIND/HASH_ADD/HASH_DEL */ +#define HASH_FIND_STR(head,findstr,out) \ + HASH_FIND(hh,head,findstr,(unsigned)uthash_strlen(findstr),out) +#define HASH_ADD_STR(head,strfield,add) \ + HASH_ADD(hh,head,strfield[0],(unsigned)uthash_strlen(add->strfield),add) +#define HASH_REPLACE_STR(head,strfield,add,replaced) \ + HASH_REPLACE(hh,head,strfield[0],(unsigned)uthash_strlen(add->strfield),add,replaced) +#define HASH_FIND_INT(head,findint,out) \ + HASH_FIND(hh,head,findint,sizeof(int),out) +#define HASH_ADD_INT(head,intfield,add) \ + HASH_ADD(hh,head,intfield,sizeof(int),add) +#define HASH_REPLACE_INT(head,intfield,add,replaced) \ + HASH_REPLACE(hh,head,intfield,sizeof(int),add,replaced) +#define HASH_FIND_PTR(head,findptr,out) \ + HASH_FIND(hh,head,findptr,sizeof(void *),out) +#define HASH_ADD_PTR(head,ptrfield,add) \ + HASH_ADD(hh,head,ptrfield,sizeof(void *),add) +#define HASH_REPLACE_PTR(head,ptrfield,add,replaced) \ + HASH_REPLACE(hh,head,ptrfield,sizeof(void *),add,replaced) +#define HASH_DEL(head,delptr) \ + HASH_DELETE(hh,head,delptr) + +/* HASH_FSCK checks hash integrity on every add/delete when HASH_DEBUG is defined. + * This is for uthash developer only; it compiles away if HASH_DEBUG isn't defined. + */ +#ifdef HASH_DEBUG +#define HASH_OOPS(...) do { fprintf(stderr,__VA_ARGS__); exit(-1); } while (0) +#define HASH_FSCK(hh,head) \ +do { \ + struct UT_hash_handle *_thh; \ + if (head) { \ + unsigned _bkt_i; \ + unsigned _count; \ + char *_prev; \ + _count = 0; \ + for( _bkt_i = 0; _bkt_i < (head)->hh.tbl->num_buckets; _bkt_i++) { \ + unsigned _bkt_count = 0; \ + _thh = (head)->hh.tbl->buckets[_bkt_i].hh_head; \ + _prev = NULL; \ + while (_thh) { \ + if (_prev != (char*)(_thh->hh_prev)) { \ + HASH_OOPS("invalid hh_prev %p, actual %p\n", \ + _thh->hh_prev, _prev ); \ + } \ + _bkt_count++; \ + _prev = (char*)(_thh); \ + _thh = _thh->hh_next; \ + } \ + _count += _bkt_count; \ + if ((head)->hh.tbl->buckets[_bkt_i].count != _bkt_count) { \ + HASH_OOPS("invalid bucket count %u, actual %u\n", \ + (head)->hh.tbl->buckets[_bkt_i].count, _bkt_count); \ + } \ + } \ + if (_count != (head)->hh.tbl->num_items) { \ + HASH_OOPS("invalid hh item count %u, actual %u\n", \ + (head)->hh.tbl->num_items, _count ); \ + } \ + /* traverse hh in app order; check next/prev integrity, count */ \ + _count = 0; \ + _prev = NULL; \ + _thh = &(head)->hh; \ + while (_thh) { \ + _count++; \ + if (_prev !=(char*)(_thh->prev)) { \ + HASH_OOPS("invalid prev %p, actual %p\n", \ + _thh->prev, _prev ); \ + } \ + _prev = (char*)ELMT_FROM_HH((head)->hh.tbl, _thh); \ + _thh = ( _thh->next ? (UT_hash_handle*)((char*)(_thh->next) + \ + (head)->hh.tbl->hho) : NULL ); \ + } \ + if (_count != (head)->hh.tbl->num_items) { \ + HASH_OOPS("invalid app item count %u, actual %u\n", \ + (head)->hh.tbl->num_items, _count ); \ + } \ + } \ +} while (0) +#else +#define HASH_FSCK(hh,head) +#endif + +/* When compiled with -DHASH_EMIT_KEYS, length-prefixed keys are emitted to + * the descriptor to which this macro is defined for tuning the hash function. + * The app can #include to get the prototype for write(2). */ +#ifdef HASH_EMIT_KEYS +#define HASH_EMIT_KEY(hh,head,keyptr,fieldlen) \ +do { \ + unsigned _klen = fieldlen; \ + write(HASH_EMIT_KEYS, &_klen, sizeof(_klen)); \ + write(HASH_EMIT_KEYS, keyptr, (unsigned long)fieldlen); \ +} while (0) +#else +#define HASH_EMIT_KEY(hh,head,keyptr,fieldlen) +#endif + +/* default to Jenkin's hash unless overridden e.g. DHASH_FUNCTION=HASH_SAX */ +#ifdef HASH_FUNCTION +#define HASH_FCN HASH_FUNCTION +#else +#define HASH_FCN HASH_JEN +#endif + +/* The Bernstein hash function, used in Perl prior to v5.6. Note (x<<5+x)=x*33. */ +#define HASH_BER(key,keylen,hashv) \ +do { \ + unsigned _hb_keylen=(unsigned)keylen; \ + const unsigned char *_hb_key=(const unsigned char*)(key); \ + (hashv) = 0; \ + while (_hb_keylen-- != 0U) { \ + (hashv) = (((hashv) << 5) + (hashv)) + *_hb_key++; \ + } \ +} while (0) + + +/* SAX/FNV/OAT/JEN hash functions are macro variants of those listed at + * http://eternallyconfuzzled.com/tuts/algorithms/jsw_tut_hashing.aspx */ +#define HASH_SAX(key,keylen,hashv) \ +do { \ + unsigned _sx_i; \ + const unsigned char *_hs_key=(const unsigned char*)(key); \ + hashv = 0; \ + for(_sx_i=0; _sx_i < keylen; _sx_i++) { \ + hashv ^= (hashv << 5) + (hashv >> 2) + _hs_key[_sx_i]; \ + } \ +} while (0) +/* FNV-1a variation */ +#define HASH_FNV(key,keylen,hashv) \ +do { \ + unsigned _fn_i; \ + const unsigned char *_hf_key=(const unsigned char*)(key); \ + hashv = 2166136261U; \ + for(_fn_i=0; _fn_i < keylen; _fn_i++) { \ + hashv = hashv ^ _hf_key[_fn_i]; \ + hashv = hashv * 16777619U; \ + } \ +} while (0) + +#define HASH_OAT(key,keylen,hashv) \ +do { \ + unsigned _ho_i; \ + const unsigned char *_ho_key=(const unsigned char*)(key); \ + hashv = 0; \ + for(_ho_i=0; _ho_i < keylen; _ho_i++) { \ + hashv += _ho_key[_ho_i]; \ + hashv += (hashv << 10); \ + hashv ^= (hashv >> 6); \ + } \ + hashv += (hashv << 3); \ + hashv ^= (hashv >> 11); \ + hashv += (hashv << 15); \ +} while (0) + +#define HASH_JEN_MIX(a,b,c) \ +do { \ + a -= b; a -= c; a ^= ( c >> 13 ); \ + b -= c; b -= a; b ^= ( a << 8 ); \ + c -= a; c -= b; c ^= ( b >> 13 ); \ + a -= b; a -= c; a ^= ( c >> 12 ); \ + b -= c; b -= a; b ^= ( a << 16 ); \ + c -= a; c -= b; c ^= ( b >> 5 ); \ + a -= b; a -= c; a ^= ( c >> 3 ); \ + b -= c; b -= a; b ^= ( a << 10 ); \ + c -= a; c -= b; c ^= ( b >> 15 ); \ +} while (0) + +#define HASH_JEN(key,keylen,hashv) \ +do { \ + unsigned _hj_i,_hj_j,_hj_k; \ + unsigned const char *_hj_key=(unsigned const char*)(key); \ + hashv = 0xfeedbeefu; \ + _hj_i = _hj_j = 0x9e3779b9u; \ + _hj_k = (unsigned)(keylen); \ + while (_hj_k >= 12U) { \ + _hj_i += (_hj_key[0] + ( (unsigned)_hj_key[1] << 8 ) \ + + ( (unsigned)_hj_key[2] << 16 ) \ + + ( (unsigned)_hj_key[3] << 24 ) ); \ + _hj_j += (_hj_key[4] + ( (unsigned)_hj_key[5] << 8 ) \ + + ( (unsigned)_hj_key[6] << 16 ) \ + + ( (unsigned)_hj_key[7] << 24 ) ); \ + hashv += (_hj_key[8] + ( (unsigned)_hj_key[9] << 8 ) \ + + ( (unsigned)_hj_key[10] << 16 ) \ + + ( (unsigned)_hj_key[11] << 24 ) ); \ + \ + HASH_JEN_MIX(_hj_i, _hj_j, hashv); \ + \ + _hj_key += 12; \ + _hj_k -= 12U; \ + } \ + hashv += (unsigned)(keylen); \ + switch ( _hj_k ) { \ + case 11: hashv += ( (unsigned)_hj_key[10] << 24 ); /* FALLTHROUGH */ \ + case 10: hashv += ( (unsigned)_hj_key[9] << 16 ); /* FALLTHROUGH */ \ + case 9: hashv += ( (unsigned)_hj_key[8] << 8 ); /* FALLTHROUGH */ \ + case 8: _hj_j += ( (unsigned)_hj_key[7] << 24 ); /* FALLTHROUGH */ \ + case 7: _hj_j += ( (unsigned)_hj_key[6] << 16 ); /* FALLTHROUGH */ \ + case 6: _hj_j += ( (unsigned)_hj_key[5] << 8 ); /* FALLTHROUGH */ \ + case 5: _hj_j += _hj_key[4]; /* FALLTHROUGH */ \ + case 4: _hj_i += ( (unsigned)_hj_key[3] << 24 ); /* FALLTHROUGH */ \ + case 3: _hj_i += ( (unsigned)_hj_key[2] << 16 ); /* FALLTHROUGH */ \ + case 2: _hj_i += ( (unsigned)_hj_key[1] << 8 ); /* FALLTHROUGH */ \ + case 1: _hj_i += _hj_key[0]; \ + } \ + HASH_JEN_MIX(_hj_i, _hj_j, hashv); \ +} while (0) + +/* The Paul Hsieh hash function */ +#undef get16bits +#if (defined(__GNUC__) && defined(__i386__)) || defined(__WATCOMC__) \ + || defined(_MSC_VER) || defined (__BORLANDC__) || defined (__TURBOC__) +#define get16bits(d) (*((const uint16_t *) (d))) +#endif + +#if !defined (get16bits) +#define get16bits(d) ((((uint32_t)(((const uint8_t *)(d))[1])) << 8) \ + +(uint32_t)(((const uint8_t *)(d))[0]) ) +#endif +#define HASH_SFH(key,keylen,hashv) \ +do { \ + unsigned const char *_sfh_key=(unsigned const char*)(key); \ + uint32_t _sfh_tmp, _sfh_len = (uint32_t)keylen; \ + \ + unsigned _sfh_rem = _sfh_len & 3U; \ + _sfh_len >>= 2; \ + hashv = 0xcafebabeu; \ + \ + /* Main loop */ \ + for (;_sfh_len > 0U; _sfh_len--) { \ + hashv += get16bits (_sfh_key); \ + _sfh_tmp = ((uint32_t)(get16bits (_sfh_key+2)) << 11) ^ hashv; \ + hashv = (hashv << 16) ^ _sfh_tmp; \ + _sfh_key += 2U*sizeof (uint16_t); \ + hashv += hashv >> 11; \ + } \ + \ + /* Handle end cases */ \ + switch (_sfh_rem) { \ + case 3: hashv += get16bits (_sfh_key); \ + hashv ^= hashv << 16; \ + hashv ^= (uint32_t)(_sfh_key[sizeof (uint16_t)]) << 18; \ + hashv += hashv >> 11; \ + break; \ + case 2: hashv += get16bits (_sfh_key); \ + hashv ^= hashv << 11; \ + hashv += hashv >> 17; \ + break; \ + case 1: hashv += *_sfh_key; \ + hashv ^= hashv << 10; \ + hashv += hashv >> 1; \ + } \ + \ + /* Force "avalanching" of final 127 bits */ \ + hashv ^= hashv << 3; \ + hashv += hashv >> 5; \ + hashv ^= hashv << 4; \ + hashv += hashv >> 17; \ + hashv ^= hashv << 25; \ + hashv += hashv >> 6; \ +} while (0) + +#ifdef HASH_USING_NO_STRICT_ALIASING +/* The MurmurHash exploits some CPU's (x86,x86_64) tolerance for unaligned reads. + * For other types of CPU's (e.g. Sparc) an unaligned read causes a bus error. + * MurmurHash uses the faster approach only on CPU's where we know it's safe. + * + * Note the preprocessor built-in defines can be emitted using: + * + * gcc -m64 -dM -E - < /dev/null (on gcc) + * cc -## a.c (where a.c is a simple test file) (Sun Studio) + */ +#if (defined(__i386__) || defined(__x86_64__) || defined(_M_IX86)) +#define MUR_GETBLOCK(p,i) p[i] +#else /* non intel */ +#define MUR_PLUS0_ALIGNED(p) (((unsigned long)p & 3UL) == 0UL) +#define MUR_PLUS1_ALIGNED(p) (((unsigned long)p & 3UL) == 1UL) +#define MUR_PLUS2_ALIGNED(p) (((unsigned long)p & 3UL) == 2UL) +#define MUR_PLUS3_ALIGNED(p) (((unsigned long)p & 3UL) == 3UL) +#define WP(p) ((uint32_t*)((unsigned long)(p) & ~3UL)) +#if (defined(__BIG_ENDIAN__) || defined(SPARC) || defined(__ppc__) || defined(__ppc64__)) +#define MUR_THREE_ONE(p) ((((*WP(p))&0x00ffffff) << 8) | (((*(WP(p)+1))&0xff000000) >> 24)) +#define MUR_TWO_TWO(p) ((((*WP(p))&0x0000ffff) <<16) | (((*(WP(p)+1))&0xffff0000) >> 16)) +#define MUR_ONE_THREE(p) ((((*WP(p))&0x000000ff) <<24) | (((*(WP(p)+1))&0xffffff00) >> 8)) +#else /* assume little endian non-intel */ +#define MUR_THREE_ONE(p) ((((*WP(p))&0xffffff00) >> 8) | (((*(WP(p)+1))&0x000000ff) << 24)) +#define MUR_TWO_TWO(p) ((((*WP(p))&0xffff0000) >>16) | (((*(WP(p)+1))&0x0000ffff) << 16)) +#define MUR_ONE_THREE(p) ((((*WP(p))&0xff000000) >>24) | (((*(WP(p)+1))&0x00ffffff) << 8)) +#endif +#define MUR_GETBLOCK(p,i) (MUR_PLUS0_ALIGNED(p) ? ((p)[i]) : \ + (MUR_PLUS1_ALIGNED(p) ? MUR_THREE_ONE(p) : \ + (MUR_PLUS2_ALIGNED(p) ? MUR_TWO_TWO(p) : \ + MUR_ONE_THREE(p)))) +#endif +#define MUR_ROTL32(x,r) (((x) << (r)) | ((x) >> (32 - (r)))) +#define MUR_FMIX(_h) \ +do { \ + _h ^= _h >> 16; \ + _h *= 0x85ebca6bu; \ + _h ^= _h >> 13; \ + _h *= 0xc2b2ae35u; \ + _h ^= _h >> 16; \ +} while (0) + +#define HASH_MUR(key,keylen,hashv) \ +do { \ + const uint8_t *_mur_data = (const uint8_t*)(key); \ + const int _mur_nblocks = (int)(keylen) / 4; \ + uint32_t _mur_h1 = 0xf88D5353u; \ + uint32_t _mur_c1 = 0xcc9e2d51u; \ + uint32_t _mur_c2 = 0x1b873593u; \ + uint32_t _mur_k1 = 0; \ + const uint8_t *_mur_tail; \ + const uint32_t *_mur_blocks = (const uint32_t*)(_mur_data+(_mur_nblocks*4)); \ + int _mur_i; \ + for(_mur_i = -_mur_nblocks; _mur_i!=0; _mur_i++) { \ + _mur_k1 = MUR_GETBLOCK(_mur_blocks,_mur_i); \ + _mur_k1 *= _mur_c1; \ + _mur_k1 = MUR_ROTL32(_mur_k1,15); \ + _mur_k1 *= _mur_c2; \ + \ + _mur_h1 ^= _mur_k1; \ + _mur_h1 = MUR_ROTL32(_mur_h1,13); \ + _mur_h1 = (_mur_h1*5U) + 0xe6546b64u; \ + } \ + _mur_tail = (const uint8_t*)(_mur_data + (_mur_nblocks*4)); \ + _mur_k1=0; \ + switch((keylen) & 3U) { \ + case 3: _mur_k1 ^= (uint32_t)_mur_tail[2] << 16; /* FALLTHROUGH */ \ + case 2: _mur_k1 ^= (uint32_t)_mur_tail[1] << 8; /* FALLTHROUGH */ \ + case 1: _mur_k1 ^= (uint32_t)_mur_tail[0]; \ + _mur_k1 *= _mur_c1; \ + _mur_k1 = MUR_ROTL32(_mur_k1,15); \ + _mur_k1 *= _mur_c2; \ + _mur_h1 ^= _mur_k1; \ + } \ + _mur_h1 ^= (uint32_t)(keylen); \ + MUR_FMIX(_mur_h1); \ + hashv = _mur_h1; \ +} while (0) +#endif /* HASH_USING_NO_STRICT_ALIASING */ + +/* iterate over items in a known bucket to find desired item */ +#define HASH_FIND_IN_BKT(tbl,hh,head,keyptr,keylen_in,hashval,out) \ +do { \ + if ((head).hh_head != NULL) { \ + DECLTYPE_ASSIGN(out, ELMT_FROM_HH(tbl, (head).hh_head)); \ + } else { \ + (out) = NULL; \ + } \ + while ((out) != NULL) { \ + if ((out)->hh.hashv == (hashval) && (out)->hh.keylen == (keylen_in)) { \ + if (uthash_memcmp((out)->hh.key, keyptr, keylen_in) == 0) { \ + break; \ + } \ + } \ + if ((out)->hh.hh_next != NULL) { \ + DECLTYPE_ASSIGN(out, ELMT_FROM_HH(tbl, (out)->hh.hh_next)); \ + } else { \ + (out) = NULL; \ + } \ + } \ +} while (0) + +/* add an item to a bucket */ +#define HASH_ADD_TO_BKT(head,addhh) \ +do { \ + head.count++; \ + (addhh)->hh_next = head.hh_head; \ + (addhh)->hh_prev = NULL; \ + if (head.hh_head != NULL) { (head).hh_head->hh_prev = (addhh); } \ + (head).hh_head=addhh; \ + if ((head.count >= ((head.expand_mult+1U) * HASH_BKT_CAPACITY_THRESH)) \ + && ((addhh)->tbl->noexpand != 1U)) { \ + HASH_EXPAND_BUCKETS((addhh)->tbl); \ + } \ +} while (0) + +/* remove an item from a given bucket */ +#define HASH_DEL_IN_BKT(hh,head,hh_del) \ + (head).count--; \ + if ((head).hh_head == hh_del) { \ + (head).hh_head = hh_del->hh_next; \ + } \ + if (hh_del->hh_prev) { \ + hh_del->hh_prev->hh_next = hh_del->hh_next; \ + } \ + if (hh_del->hh_next) { \ + hh_del->hh_next->hh_prev = hh_del->hh_prev; \ + } + +/* Bucket expansion has the effect of doubling the number of buckets + * and redistributing the items into the new buckets. Ideally the + * items will distribute more or less evenly into the new buckets + * (the extent to which this is true is a measure of the quality of + * the hash function as it applies to the key domain). + * + * With the items distributed into more buckets, the chain length + * (item count) in each bucket is reduced. Thus by expanding buckets + * the hash keeps a bound on the chain length. This bounded chain + * length is the essence of how a hash provides constant time lookup. + * + * The calculation of tbl->ideal_chain_maxlen below deserves some + * explanation. First, keep in mind that we're calculating the ideal + * maximum chain length based on the *new* (doubled) bucket count. + * In fractions this is just n/b (n=number of items,b=new num buckets). + * Since the ideal chain length is an integer, we want to calculate + * ceil(n/b). We don't depend on floating point arithmetic in this + * hash, so to calculate ceil(n/b) with integers we could write + * + * ceil(n/b) = (n/b) + ((n%b)?1:0) + * + * and in fact a previous version of this hash did just that. + * But now we have improved things a bit by recognizing that b is + * always a power of two. We keep its base 2 log handy (call it lb), + * so now we can write this with a bit shift and logical AND: + * + * ceil(n/b) = (n>>lb) + ( (n & (b-1)) ? 1:0) + * + */ +#define HASH_EXPAND_BUCKETS(tbl) \ +do { \ + unsigned _he_bkt; \ + unsigned _he_bkt_i; \ + struct UT_hash_handle *_he_thh, *_he_hh_nxt; \ + UT_hash_bucket *_he_new_buckets, *_he_newbkt; \ + _he_new_buckets = (UT_hash_bucket*)uthash_malloc( \ + 2UL * tbl->num_buckets * sizeof(struct UT_hash_bucket)); \ + if (!_he_new_buckets) { uthash_fatal( "out of memory"); } \ + memset(_he_new_buckets, 0, \ + 2UL * tbl->num_buckets * sizeof(struct UT_hash_bucket)); \ + tbl->ideal_chain_maxlen = \ + (tbl->num_items >> (tbl->log2_num_buckets+1U)) + \ + (((tbl->num_items & ((tbl->num_buckets*2U)-1U)) != 0U) ? 1U : 0U); \ + tbl->nonideal_items = 0; \ + for(_he_bkt_i = 0; _he_bkt_i < tbl->num_buckets; _he_bkt_i++) \ + { \ + _he_thh = tbl->buckets[ _he_bkt_i ].hh_head; \ + while (_he_thh != NULL) { \ + _he_hh_nxt = _he_thh->hh_next; \ + HASH_TO_BKT( _he_thh->hashv, tbl->num_buckets*2U, _he_bkt); \ + _he_newbkt = &(_he_new_buckets[ _he_bkt ]); \ + if (++(_he_newbkt->count) > tbl->ideal_chain_maxlen) { \ + tbl->nonideal_items++; \ + _he_newbkt->expand_mult = _he_newbkt->count / \ + tbl->ideal_chain_maxlen; \ + } \ + _he_thh->hh_prev = NULL; \ + _he_thh->hh_next = _he_newbkt->hh_head; \ + if (_he_newbkt->hh_head != NULL) { _he_newbkt->hh_head->hh_prev = \ + _he_thh; } \ + _he_newbkt->hh_head = _he_thh; \ + _he_thh = _he_hh_nxt; \ + } \ + } \ + uthash_free( tbl->buckets, tbl->num_buckets*sizeof(struct UT_hash_bucket) ); \ + tbl->num_buckets *= 2U; \ + tbl->log2_num_buckets++; \ + tbl->buckets = _he_new_buckets; \ + tbl->ineff_expands = (tbl->nonideal_items > (tbl->num_items >> 1)) ? \ + (tbl->ineff_expands+1U) : 0U; \ + if (tbl->ineff_expands > 1U) { \ + tbl->noexpand=1; \ + uthash_noexpand_fyi(tbl); \ + } \ + uthash_expand_fyi(tbl); \ +} while (0) + + +/* This is an adaptation of Simon Tatham's O(n log(n)) mergesort */ +/* Note that HASH_SORT assumes the hash handle name to be hh. + * HASH_SRT was added to allow the hash handle name to be passed in. */ +#define HASH_SORT(head,cmpfcn) HASH_SRT(hh,head,cmpfcn) +#define HASH_SRT(hh,head,cmpfcn) \ +do { \ + unsigned _hs_i; \ + unsigned _hs_looping,_hs_nmerges,_hs_insize,_hs_psize,_hs_qsize; \ + struct UT_hash_handle *_hs_p, *_hs_q, *_hs_e, *_hs_list, *_hs_tail; \ + if (head != NULL) { \ + _hs_insize = 1; \ + _hs_looping = 1; \ + _hs_list = &((head)->hh); \ + while (_hs_looping != 0U) { \ + _hs_p = _hs_list; \ + _hs_list = NULL; \ + _hs_tail = NULL; \ + _hs_nmerges = 0; \ + while (_hs_p != NULL) { \ + _hs_nmerges++; \ + _hs_q = _hs_p; \ + _hs_psize = 0; \ + for ( _hs_i = 0; _hs_i < _hs_insize; _hs_i++ ) { \ + _hs_psize++; \ + _hs_q = (UT_hash_handle*)((_hs_q->next != NULL) ? \ + ((void*)((char*)(_hs_q->next) + \ + (head)->hh.tbl->hho)) : NULL); \ + if (! (_hs_q) ) { break; } \ + } \ + _hs_qsize = _hs_insize; \ + while ((_hs_psize > 0U) || ((_hs_qsize > 0U) && (_hs_q != NULL))) {\ + if (_hs_psize == 0U) { \ + _hs_e = _hs_q; \ + _hs_q = (UT_hash_handle*)((_hs_q->next != NULL) ? \ + ((void*)((char*)(_hs_q->next) + \ + (head)->hh.tbl->hho)) : NULL); \ + _hs_qsize--; \ + } else if ( (_hs_qsize == 0U) || (_hs_q == NULL) ) { \ + _hs_e = _hs_p; \ + if (_hs_p != NULL){ \ + _hs_p = (UT_hash_handle*)((_hs_p->next != NULL) ? \ + ((void*)((char*)(_hs_p->next) + \ + (head)->hh.tbl->hho)) : NULL); \ + } \ + _hs_psize--; \ + } else if (( \ + cmpfcn(DECLTYPE(head)(ELMT_FROM_HH((head)->hh.tbl,_hs_p)), \ + DECLTYPE(head)(ELMT_FROM_HH((head)->hh.tbl,_hs_q))) \ + ) <= 0) { \ + _hs_e = _hs_p; \ + if (_hs_p != NULL){ \ + _hs_p = (UT_hash_handle*)((_hs_p->next != NULL) ? \ + ((void*)((char*)(_hs_p->next) + \ + (head)->hh.tbl->hho)) : NULL); \ + } \ + _hs_psize--; \ + } else { \ + _hs_e = _hs_q; \ + _hs_q = (UT_hash_handle*)((_hs_q->next != NULL) ? \ + ((void*)((char*)(_hs_q->next) + \ + (head)->hh.tbl->hho)) : NULL); \ + _hs_qsize--; \ + } \ + if ( _hs_tail != NULL ) { \ + _hs_tail->next = ((_hs_e != NULL) ? \ + ELMT_FROM_HH((head)->hh.tbl,_hs_e) : NULL); \ + } else { \ + _hs_list = _hs_e; \ + } \ + if (_hs_e != NULL) { \ + _hs_e->prev = ((_hs_tail != NULL) ? \ + ELMT_FROM_HH((head)->hh.tbl,_hs_tail) : NULL); \ + } \ + _hs_tail = _hs_e; \ + } \ + _hs_p = _hs_q; \ + } \ + if (_hs_tail != NULL){ \ + _hs_tail->next = NULL; \ + } \ + if ( _hs_nmerges <= 1U ) { \ + _hs_looping=0; \ + (head)->hh.tbl->tail = _hs_tail; \ + DECLTYPE_ASSIGN(head,ELMT_FROM_HH((head)->hh.tbl, _hs_list)); \ + } \ + _hs_insize *= 2U; \ + } \ + HASH_FSCK(hh,head); \ + } \ +} while (0) + +/* This function selects items from one hash into another hash. + * The end result is that the selected items have dual presence + * in both hashes. There is no copy of the items made; rather + * they are added into the new hash through a secondary hash + * hash handle that must be present in the structure. */ +#define HASH_SELECT(hh_dst, dst, hh_src, src, cond) \ +do { \ + unsigned _src_bkt, _dst_bkt; \ + void *_last_elt=NULL, *_elt; \ + UT_hash_handle *_src_hh, *_dst_hh, *_last_elt_hh=NULL; \ + ptrdiff_t _dst_hho = ((char*)(&(dst)->hh_dst) - (char*)(dst)); \ + if (src != NULL) { \ + for(_src_bkt=0; _src_bkt < (src)->hh_src.tbl->num_buckets; _src_bkt++) { \ + for(_src_hh = (src)->hh_src.tbl->buckets[_src_bkt].hh_head; \ + _src_hh != NULL; \ + _src_hh = _src_hh->hh_next) { \ + _elt = ELMT_FROM_HH((src)->hh_src.tbl, _src_hh); \ + if (cond(_elt)) { \ + _dst_hh = (UT_hash_handle*)(((char*)_elt) + _dst_hho); \ + _dst_hh->key = _src_hh->key; \ + _dst_hh->keylen = _src_hh->keylen; \ + _dst_hh->hashv = _src_hh->hashv; \ + _dst_hh->prev = _last_elt; \ + _dst_hh->next = NULL; \ + if (_last_elt_hh != NULL) { _last_elt_hh->next = _elt; } \ + if (dst == NULL) { \ + DECLTYPE_ASSIGN(dst,_elt); \ + HASH_MAKE_TABLE(hh_dst,dst); \ + } else { \ + _dst_hh->tbl = (dst)->hh_dst.tbl; \ + } \ + HASH_TO_BKT(_dst_hh->hashv, _dst_hh->tbl->num_buckets, _dst_bkt); \ + HASH_ADD_TO_BKT(_dst_hh->tbl->buckets[_dst_bkt],_dst_hh); \ + (dst)->hh_dst.tbl->num_items++; \ + _last_elt = _elt; \ + _last_elt_hh = _dst_hh; \ + } \ + } \ + } \ + } \ + HASH_FSCK(hh_dst,dst); \ +} while (0) + +#define HASH_CLEAR(hh,head) \ +do { \ + if (head != NULL) { \ + uthash_free((head)->hh.tbl->buckets, \ + (head)->hh.tbl->num_buckets*sizeof(struct UT_hash_bucket)); \ + HASH_BLOOM_FREE((head)->hh.tbl); \ + uthash_free((head)->hh.tbl, sizeof(UT_hash_table)); \ + (head)=NULL; \ + } \ +} while (0) + +#define HASH_OVERHEAD(hh,head) \ + ((head != NULL) ? ( \ + (size_t)(((head)->hh.tbl->num_items * sizeof(UT_hash_handle)) + \ + ((head)->hh.tbl->num_buckets * sizeof(UT_hash_bucket)) + \ + sizeof(UT_hash_table) + \ + (HASH_BLOOM_BYTELEN))) : 0U) + +#ifdef NO_DECLTYPE +#define HASH_ITER(hh,head,el,tmp) \ +for(((el)=(head)), ((*(char**)(&(tmp)))=(char*)((head!=NULL)?(head)->hh.next:NULL)); \ + (el) != NULL; ((el)=(tmp)), ((*(char**)(&(tmp)))=(char*)((tmp!=NULL)?(tmp)->hh.next:NULL))) +#else +#define HASH_ITER(hh,head,el,tmp) \ +for(((el)=(head)), ((tmp)=DECLTYPE(el)((head!=NULL)?(head)->hh.next:NULL)); \ + (el) != NULL; ((el)=(tmp)), ((tmp)=DECLTYPE(el)((tmp!=NULL)?(tmp)->hh.next:NULL))) +#endif + +/* obtain a count of items in the hash */ +#define HASH_COUNT(head) HASH_CNT(hh,head) +#define HASH_CNT(hh,head) ((head != NULL)?((head)->hh.tbl->num_items):0U) + +typedef struct UT_hash_bucket { + struct UT_hash_handle *hh_head; + unsigned count; + + /* expand_mult is normally set to 0. In this situation, the max chain length + * threshold is enforced at its default value, HASH_BKT_CAPACITY_THRESH. (If + * the bucket's chain exceeds this length, bucket expansion is triggered). + * However, setting expand_mult to a non-zero value delays bucket expansion + * (that would be triggered by additions to this particular bucket) + * until its chain length reaches a *multiple* of HASH_BKT_CAPACITY_THRESH. + * (The multiplier is simply expand_mult+1). The whole idea of this + * multiplier is to reduce bucket expansions, since they are expensive, in + * situations where we know that a particular bucket tends to be overused. + * It is better to let its chain length grow to a longer yet-still-bounded + * value, than to do an O(n) bucket expansion too often. + */ + unsigned expand_mult; + +} UT_hash_bucket; + +/* random signature used only to find hash tables in external analysis */ +#define HASH_SIGNATURE 0xa0111fe1u +#define HASH_BLOOM_SIGNATURE 0xb12220f2u + +typedef struct UT_hash_table { + UT_hash_bucket *buckets; + unsigned num_buckets, log2_num_buckets; + unsigned num_items; + struct UT_hash_handle *tail; /* tail hh in app order, for fast append */ + ptrdiff_t hho; /* hash handle offset (byte pos of hash handle in element */ + + /* in an ideal situation (all buckets used equally), no bucket would have + * more than ceil(#items/#buckets) items. that's the ideal chain length. */ + unsigned ideal_chain_maxlen; + + /* nonideal_items is the number of items in the hash whose chain position + * exceeds the ideal chain maxlen. these items pay the penalty for an uneven + * hash distribution; reaching them in a chain traversal takes >ideal steps */ + unsigned nonideal_items; + + /* ineffective expands occur when a bucket doubling was performed, but + * afterward, more than half the items in the hash had nonideal chain + * positions. If this happens on two consecutive expansions we inhibit any + * further expansion, as it's not helping; this happens when the hash + * function isn't a good fit for the key domain. When expansion is inhibited + * the hash will still work, albeit no longer in constant time. */ + unsigned ineff_expands, noexpand; + + uint32_t signature; /* used only to find hash tables in external analysis */ +#ifdef HASH_BLOOM + uint32_t bloom_sig; /* used only to test bloom exists in external analysis */ + uint8_t *bloom_bv; + uint8_t bloom_nbits; +#endif + +} UT_hash_table; + +typedef struct UT_hash_handle { + struct UT_hash_table *tbl; + void *prev; /* prev element in app order */ + void *next; /* next element in app order */ + struct UT_hash_handle *hh_prev; /* previous hh in bucket order */ + struct UT_hash_handle *hh_next; /* next hh in bucket order */ + void *key; /* ptr to enclosing struct's key */ + unsigned keylen; /* enclosing struct's key len */ + unsigned hashv; /* result of hash-fcn(key) */ +} UT_hash_handle; + +#endif /* UTHASH_H */ diff --git a/lib/libdrm/libsync.h b/lib/libdrm/libsync.h new file mode 100644 index 000000000..f1a2f96d3 --- /dev/null +++ b/lib/libdrm/libsync.h @@ -0,0 +1,148 @@ +/* + * sync abstraction + * Copyright 2015-2016 Collabora Ltd. + * + * Based on the implementation from the Android Open Source Project, + * + * Copyright 2012 Google, Inc + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef _LIBSYNC_H +#define _LIBSYNC_H + +#include +#include +#include +#include +#include +#include +#include + +#if defined(__cplusplus) +extern "C" { +#endif + +#ifndef SYNC_IOC_MERGE +/* duplicated from linux/sync_file.h to avoid build-time dependency + * on new (v4.7) kernel headers. Once distro's are mostly using + * something newer than v4.7 drop this and #include + * instead. + */ +struct sync_merge_data { + char name[32]; + int32_t fd2; + int32_t fence; + uint32_t flags; + uint32_t pad; +}; +#define SYNC_IOC_MAGIC '>' +#define SYNC_IOC_MERGE _IOWR(SYNC_IOC_MAGIC, 3, struct sync_merge_data) +#endif + + +static inline int sync_wait(int fd, int timeout) +{ + struct pollfd fds = {0}; + int ret; + + fds.fd = fd; + fds.events = POLLIN; + + do { + ret = poll(&fds, 1, timeout); + if (ret > 0) { + if (fds.revents & (POLLERR | POLLNVAL)) { + errno = EINVAL; + return -1; + } + return 0; + } else if (ret == 0) { + errno = ETIME; + return -1; + } + } while (ret == -1 && (errno == EINTR || errno == EAGAIN)); + + return ret; +} + +static inline int sync_merge(const char *name, int fd1, int fd2) +{ + struct sync_merge_data data = {0}; + int ret; + + data.fd2 = fd2; + strncpy(data.name, name, sizeof(data.name)); + + do { + ret = ioctl(fd1, SYNC_IOC_MERGE, &data); + } while (ret == -1 && (errno == EINTR || errno == EAGAIN)); + + if (ret < 0) + return ret; + + return data.fence; +} + +/* accumulate fd2 into fd1. If *fd1 is not a valid fd then dup fd2, + * otherwise sync_merge() and close the old *fd1. This can be used + * to implement the pattern: + * + * init() + * { + * batch.fence_fd = -1; + * } + * + * // does *NOT* take ownership of fd + * server_sync(int fd) + * { + * if (sync_accumulate("foo", &batch.fence_fd, fd)) { + * ... error ... + * } + * } + */ +static inline int sync_accumulate(const char *name, int *fd1, int fd2) +{ + int ret; + + assert(fd2 >= 0); + + if (*fd1 < 0) { + *fd1 = dup(fd2); + return 0; + } + + ret = sync_merge(name, *fd1, fd2); + if (ret < 0) { + /* leave *fd1 as it is */ + return ret; + } + + close(*fd1); + *fd1 = ret; + + return 0; +} + +#if defined(__cplusplus) +} +#endif + +#endif