Signed-off-by: Tom St Denis <tom.stde...@amd.com> (v2): Use iova debugfs for read/write not address translation --- src/lib/CMakeLists.txt | 1 - src/lib/close_asic.c | 2 +- src/lib/discover.c | 3 + src/lib/free_maps.c | 44 ---------- src/lib/read_vram.c | 213 ++++--------------------------------------------- src/umr.h | 14 +--- 6 files changed, 23 insertions(+), 254 deletions(-) delete mode 100644 src/lib/free_maps.c
diff --git a/src/lib/CMakeLists.txt b/src/lib/CMakeLists.txt index 78d827ac1bf1..a21fdf8eea2d 100644 --- a/src/lib/CMakeLists.txt +++ b/src/lib/CMakeLists.txt @@ -14,7 +14,6 @@ add_library(umrcore STATIC discover.c dump_ib.c find_reg.c - free_maps.c mmio.c query_drm.c read_sensor.c diff --git a/src/lib/close_asic.c b/src/lib/close_asic.c index d532a11fa671..a140409e617b 100644 --- a/src/lib/close_asic.c +++ b/src/lib/close_asic.c @@ -29,7 +29,6 @@ void umr_free_asic(struct umr_asic *asic) { int x; - umr_free_maps(asic); if (asic->pci.mem != NULL) { // free PCI mapping pci_device_unmap_range(asic->pci.pdevice, asic->pci.mem, asic->pci.pdevice->regions[asic->pci.region].size); @@ -57,6 +56,7 @@ void umr_close_asic(struct umr_asic *asic) cond_close(asic->fd.vram); cond_close(asic->fd.gpr); cond_close(asic->fd.drm); + cond_close(asic->fd.iova); umr_free_asic(asic); } } diff --git a/src/lib/discover.c b/src/lib/discover.c index dcc212fc39e4..ff7950e4e6ba 100644 --- a/src/lib/discover.c +++ b/src/lib/discover.c @@ -222,6 +222,8 @@ struct umr_asic *umr_discover_asic(struct umr_options *options) asic->fd.vram = open(fname, O_RDWR); snprintf(fname, sizeof(fname)-1, "/sys/kernel/debug/dri/%d/amdgpu_gpr", asic->instance); asic->fd.gpr = open(fname, O_RDWR); + snprintf(fname, sizeof(fname)-1, "/sys/kernel/debug/dri/%d/amdgpu_iova", asic->instance); + asic->fd.iova = open(fname, O_RDWR); asic->fd.drm = -1; // default to closed // if appending to the fd list remember to update close_asic() and discover_by_did()... } else { @@ -235,6 +237,7 @@ struct umr_asic *umr_discover_asic(struct umr_options *options) asic->fd.vram = -1; asic->fd.gpr = -1; asic->fd.drm = -1; + asic->fd.iova = -1; } if (options->use_pci) { diff --git a/src/lib/free_maps.c b/src/lib/free_maps.c deleted file mode 100644 index e1d27cb177f7..000000000000 --- a/src/lib/free_maps.c +++ /dev/null @@ -1,44 +0,0 @@ -/* - * Copyright 2017 Advanced Micro Devices, 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 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. - * - * Authors: Tom St Denis <tom.stde...@amd.com> - * - */ -#include "umr.h" - -static void recurse_free(struct umr_map *map) -{ - if (map->left) - recurse_free(map->left); - if (map->right) - recurse_free(map->right); - free(map); -} - -void umr_free_maps(struct umr_asic *asic) -{ - if (!asic->maps) - return; - - recurse_free(asic->maps->maps); - free(asic->maps); - asic->maps = NULL; -} diff --git a/src/lib/read_vram.c b/src/lib/read_vram.c index 6e8f1f931895..928467a0a3a8 100644 --- a/src/lib/read_vram.c +++ b/src/lib/read_vram.c @@ -25,163 +25,6 @@ #include "umrapp.h" #include <inttypes.h> -// find a mapping or create node for it -static struct umr_map *find_map(struct umr_dma_maps *maps, uint64_t dma_addr, int create) -{ - struct umr_map *n = maps->maps, **nn; - uint64_t key; - - // addresses aren't terribly random - // so if we use an identity function on the search - // key we'll end up with a really unbalanced tree - // so mix up address a bit to randomize keys - key = dma_addr ^ (dma_addr >> 11); - - if (!n) { - maps->maps = calloc(1, sizeof(maps->maps[0])); - maps->maps->key = key; - return maps->maps; - } - - while (n->dma_addr != dma_addr) { - if (key > n->key) - nn = &n->left; - else - nn = &n->right; - - if (*nn) { - n = *nn; - } else { - if (!create) return NULL; - - // add the new node - *nn = calloc(1, sizeof(maps->maps[0])); - (*nn)->key = key; - return *nn; - } - } - - return n; -} - -// insert/replace mapping in array -static int insert_map(struct umr_dma_maps *maps, - uint64_t dma_addr, uint64_t phys_addr, int valid) -{ - struct umr_map *map = find_map(maps, dma_addr, valid); - - // don't add a new node if we're marking it invalid - if (map) { - if (valid) { - map->dma_addr = dma_addr; - map->phys_addr = phys_addr; - map->valid = valid; - } else { - struct umr_map *tmap = NULL; - - // if marking invalid see if we can prune - // the tree a little if the node we're marking - // as invalid only has one child - if (map->left == NULL && map->right) { - tmap = map->right; - *map = *(map->right); - } else if (map->right == NULL && map->left) { - tmap = map->left; - *map = *(map->left); - } - if (tmap) - free(tmap); - } - } - - return 0; -} - -static int check_trace = 0; - -// try to convert a DMA address to physical via trace -static uint64_t dma_to_phys(struct umr_asic *asic, uint64_t dma_addr) -{ - struct umr_map *map = find_map(asic->maps, dma_addr, 0); - - if (map == NULL) - return dma_addr; - - if (map->valid) - return map->phys_addr; - else - return map->dma_addr; -} - -static int parse_trace(struct umr_asic *asic) -{ - FILE *f; - uint64_t d, p; - char *s, buf[512]; - int err = -1, valid; - struct umr_dma_maps *maps = asic->maps; - - if (!check_trace) { - check_trace = 1; - f = fopen("/sys/kernel/debug/tracing/events/ttm/ttm_dma_map/enable", "r"); - if (!f) { - fprintf(stderr, "[WARNING]: kernel does not support TTM mapping trace, please update kernel\n"); - } else { - fgets(buf, sizeof(buf)-1, f); - if (sscanf(buf, "%"SCNu64, &d) == 1) { - if (d != 1) { - fprintf(stderr, - "[WARNING]: ttm_dma_map trace is not enabled, VM decoding may fail!\n" - "[WARNING]: Enable with: 'echo 1 > /sys/kernel/debug/tracing/events/ttm/ttm_dma_map/enable'\n" - "[WARNING]: Enable with: 'echo 1 > /sys/kernel/debug/tracing/events/ttm/ttm_dma_unmap/enable'\n"); - } - } else { - fprintf(stderr, "[ERROR]: could not read ttm_dma_map enable file\n"); - } - fclose(f); - } - } - - // try to open ~/trace first - snprintf(buf, sizeof(buf)-1, "%s/trace", getenv("HOME")); - f = fopen(buf, "r"); - if (!f) - f = fopen("/sys/kernel/debug/tracing/trace", "r"); - if (!f) - goto error; - - while (fgets(buf, sizeof(buf)-1, f)) { - valid = -1; - - s = strstr(buf, "ttm_dma_map"); - if (s) { - s += strlen("ttm_dma_map"); - valid = 1; - } else { - s = strstr(buf, "ttm_dma_unmap"); - if (s) { - s += strlen("ttm_dma_unmap"); - valid = 0; - } - } - - if (valid != -1) { - s = strstr(s, asic->options.pci.name); - if (s) { - s += strlen(asic->options.pci.name) + 2; - if (sscanf(s, "0x%"SCNx64" => 0x%"SCNx64, &d, &p) == 2) { - if (insert_map(maps, d, p, valid)) - goto error; - } - } - } - } - err = 0; -error: - fclose(f); - return err; -} - static void access_vram_via_mmio(struct umr_asic *asic, uint64_t address, uint32_t size, void *dst, int write_en) { uint32_t MM_INDEX, MM_INDEX_HI, MM_DATA; @@ -217,33 +60,22 @@ static void access_vram_via_mmio(struct umr_asic *asic, uint64_t address, uint32 #define DEBUG(...) #endif -static int umr_access_sram(uint64_t address, uint32_t size, void *dst, int write_en) +static int umr_access_sram(struct umr_asic *asic, uint64_t address, uint32_t size, void *dst, int write_en) { - int fd; - DEBUG("Reading physical sys addr: 0x%llx\n", (unsigned long long)address); - fd = open("/dev/fmem", O_RDWR); - if (fd < 0) - fd = open("/dev/mem", O_RDWR | O_DSYNC); - if (fd >= 0) { - lseek(fd, address, SEEK_SET); - if (write_en == 0) { - memset(dst, 0xFF, size); - if (read(fd, dst, size) != size) { - close(fd); - return -1; - } - } else { - if (write(fd, dst, size) != size) { - close(fd); - return -1; - } + lseek(asic->fd.iova, address, SEEK_SET); + if (write_en == 0) { + memset(dst, 0xFF, size); + if (read(asic->fd.iova, dst, size) != size) { + return -1; + } + } else { + if (write(asic->fd.iova, dst, size) != size) { + return -1; } - close(fd); - return 0; } - return -1; + return 0; } @@ -389,7 +221,7 @@ static int umr_access_vram_vi(struct umr_asic *asic, uint32_t vmid, return -1; // compute starting address - start_addr = dma_to_phys(asic, pte_fields.page_base_addr) + (address & 0xFFF); + start_addr = pte_fields.page_base_addr + (address & 0xFFF); } else { // depth == 0 == PTE only pte_idx = (address >> 12); @@ -414,7 +246,7 @@ static int umr_access_vram_vi(struct umr_asic *asic, uint32_t vmid, return -1; // compute starting address - start_addr = dma_to_phys(asic, pte_fields.page_base_addr) + (address & 0xFFF); + start_addr = pte_fields.page_base_addr + (address & 0xFFF); } // read upto 4K from it @@ -428,7 +260,7 @@ static int umr_access_vram_vi(struct umr_asic *asic, uint32_t vmid, // allow destination to be NULL to simply use decoder if (pdst) { if (pte_fields.system) { - if (umr_access_sram(start_addr, chunk_size, pdst, write_en) < 0) { + if (umr_access_sram(asic, start_addr, chunk_size, pdst, write_en) < 0) { fprintf(stderr, "[ERROR]: Cannot access system ram, perhaps CONFIG_STRICT_DEVMEM is set in your kernel config?\n"); fprintf(stderr, "[ERROR]: Alternatively download and install /dev/fmem\n"); return -1; @@ -707,7 +539,7 @@ static int umr_access_vram_ai(struct umr_asic *asic, uint32_t vmid, return -1; // compute starting address - start_addr = dma_to_phys(asic, pte_fields.page_base_addr) + (address & 0xFFF); + start_addr = pte_fields.page_base_addr + (address & 0xFFF); DEBUG("phys address to read from: %llx\n\n\n", (unsigned long long)start_addr); } else { // in AI+ the BASE_ADDR is treated like a PDE entry... @@ -756,7 +588,7 @@ static int umr_access_vram_ai(struct umr_asic *asic, uint32_t vmid, return -1; // compute starting address - start_addr = dma_to_phys(asic, pte_fields.page_base_addr) + (address & 0xFFF); + start_addr = pte_fields.page_base_addr + (address & 0xFFF); } // read upto 4K from it @@ -772,7 +604,7 @@ static int umr_access_vram_ai(struct umr_asic *asic, uint32_t vmid, // allow destination to be NULL to simply use decoder if (pdst) { if (pte_fields.system) { - if (umr_access_sram(start_addr, chunk_size, pdst, write_en) < 0) { + if (umr_access_sram(asic, start_addr, chunk_size, pdst, write_en) < 0) { fprintf(stderr, "[ERROR]: Cannot access system ram, perhaps CONFIG_STRICT_DEVMEM is set in your kernel config?\n"); fprintf(stderr, "[ERROR]: Alternatively download and install /dev/fmem\n"); return -1; @@ -825,17 +657,6 @@ int umr_access_vram(struct umr_asic *asic, uint32_t vmid, uint64_t address, uint return 0; } - if (!asic->maps) { - asic->maps = calloc(1, sizeof(asic->maps[0])); - if (!asic->maps) { - fprintf(stderr, "[ERROR]: Out of memory building dma maps\n"); - return -1; - } - - if (parse_trace(asic)) - return -1; - } - switch (asic->family) { case FAMILY_SI: case FAMILY_CIK: diff --git a/src/umr.h b/src/umr.h index 3d2252e35608..a374a0d23301 100644 --- a/src/umr.h +++ b/src/umr.h @@ -210,17 +210,6 @@ struct umr_options { } pci; }; -struct umr_dma_maps { - struct umr_map { - uint64_t - dma_addr, - phys_addr, - key; - int valid; - struct umr_map *left, *right; - } *maps; -}; - struct umr_asic { char *asicname; int no_blocks; @@ -243,7 +232,8 @@ struct umr_asic { drm, wave, vram, - gpr; + gpr, + iova; } fd; struct { struct pci_device *pdevice; -- 2.12.0 _______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx