File size: 5,135 Bytes
c9c3262 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 | #include "backend-dispatched.h"
#include "backend-virgl-apir.h"
#include "ggml-backend-impl.h"
#include "ggml-backend.h"
#include "ggml-impl.h"
#include <cstdint>
static uint32_t validate_buffer_operation(size_t offset, size_t size, const char * operation) {
// Only check for critical integer overflow - no arbitrary size limits
if (offset > SIZE_MAX - size) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Integer overflow in offset+size: %zu + %zu\n", operation, offset, size);
return 1;
}
return 0; // Valid
}
uint32_t backend_buffer_get_base(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
GGML_UNUSED(ctx);
ggml_backend_buffer_t buffer;
buffer = apir_decode_ggml_buffer(dec);
if (!buffer || apir_decoder_get_fatal(dec)) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Invalid buffer handle from guest\n", __func__);
return 1;
}
uintptr_t base = (uintptr_t) buffer->iface.get_base(buffer);
apir_encode_uintptr_t(enc, &base);
return 0;
}
uint32_t backend_buffer_set_tensor(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
GGML_UNUSED(ctx);
GGML_UNUSED(enc);
ggml_backend_buffer_t buffer;
buffer = apir_decode_ggml_buffer(dec);
if (!buffer || apir_decoder_get_fatal(dec)) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Invalid buffer handle from guest\n", __func__);
return 1;
}
ggml_tensor * tensor;
// safe to remove the const qualifier here
tensor = (ggml_tensor *) (uintptr_t) apir_decode_ggml_tensor(dec);
uint32_t shmem_res_id;
apir_decode_virtgpu_shmem_res_id(dec, &shmem_res_id);
size_t offset;
apir_decode_size_t(dec, &offset);
size_t size;
apir_decode_size_t(dec, &size);
if (validate_buffer_operation(offset, size, __func__) != 0) {
return 1;
}
void * shmem_data = ctx->iface->get_shmem_ptr(ctx->ctx_id, shmem_res_id);
if (!shmem_data) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Couldn't get the shmem addr from virgl\n", __func__);
return 1;
}
buffer->iface.set_tensor(buffer, tensor, shmem_data, offset, size);
return 0;
}
uint32_t backend_buffer_get_tensor(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
GGML_UNUSED(ctx);
GGML_UNUSED(enc);
ggml_backend_buffer_t buffer;
buffer = apir_decode_ggml_buffer(dec);
if (!buffer || apir_decoder_get_fatal(dec)) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Invalid buffer handle from guest\n", __func__);
return 1;
}
const ggml_tensor * tensor;
// safe to remove the const qualifier here
tensor = apir_decode_ggml_tensor(dec);
uint32_t shmem_res_id;
apir_decode_virtgpu_shmem_res_id(dec, &shmem_res_id);
size_t offset;
apir_decode_size_t(dec, &offset);
size_t size;
apir_decode_size_t(dec, &size);
if (validate_buffer_operation(offset, size, __func__) != 0) {
return 1;
}
void * shmem_data = ctx->iface->get_shmem_ptr(ctx->ctx_id, shmem_res_id);
if (!shmem_data) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Couldn't get the shmem addr from virgl\n", __func__);
return 1;
}
buffer->iface.get_tensor(buffer, tensor, shmem_data, offset, size);
return 0;
}
uint32_t backend_buffer_cpy_tensor(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
GGML_UNUSED(ctx);
ggml_backend_buffer_t buffer;
buffer = apir_decode_ggml_buffer(dec);
if (!buffer || apir_decoder_get_fatal(dec)) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Invalid buffer handle from guest\n", __func__);
return 1;
}
const ggml_tensor * src;
// safe to remove the const qualifier here
src = apir_decode_ggml_tensor(dec);
ggml_tensor * dst = (ggml_tensor *) (uintptr_t) apir_decode_ggml_tensor(dec);
bool ret = buffer->iface.cpy_tensor(buffer, src, (ggml_tensor *) dst);
apir_encode_bool_t(enc, &ret);
return 0;
}
uint32_t backend_buffer_clear(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
GGML_UNUSED(ctx);
GGML_UNUSED(enc);
ggml_backend_buffer_t buffer;
buffer = apir_decode_ggml_buffer(dec);
if (!buffer || apir_decoder_get_fatal(dec)) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Invalid buffer handle from guest\n", __func__);
return 1;
}
uint8_t value;
apir_decode_uint8_t(dec, &value);
buffer->iface.clear(buffer, value);
return 0;
}
uint32_t backend_buffer_free_buffer(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
GGML_UNUSED(ctx);
GGML_UNUSED(enc);
ggml_backend_buffer_t buffer;
buffer = apir_decode_ggml_buffer(dec);
if (!buffer || apir_decoder_get_fatal(dec)) {
GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Invalid buffer handle from guest\n", __func__);
return 1;
}
if (!apir_untrack_backend_buffer(buffer)) {
GGML_LOG_WARN(GGML_VIRTGPU_BCK "%s: unknown buffer %p\n", __func__, (void *) buffer);
return 1;
}
buffer->iface.free_buffer(buffer);
return 0;
}
|