Superminaren/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-01 05:56:52 +02:00
Code Issues Packages Projects Releases Wiki Activity

gguf : reject tensor size that wraps after padding (#26979)

Browse Source 
GGML_PAD(nbytes, alignment) wraps to 0 when nbytes is within
(alignment - 1) of SIZE_MAX, which silently bypassed the size
overflow guard in gguf_init_from_reader. Reject the tensor before
padding when nbytes + (alignment - 1) would overflow.

Adds a test-gguf handcrafted case (F32, ne = [4, 2^30-1, 2^30+1, 1])
whose ggml_nbytes = 2^64 - 16 lands in the wrap window. Fails on
master, passes with the guard.
This commit is contained in:
Xiang Chen
2026-09-29 22:46:20 +02:00
committed by GitHub
parent d3954b9324
commit a6ea155d3d
2 changed files with 20 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/gguf.cpp
+10 -1
View File
@@ -792,7 +792,16 @@ static struct gguf_context * gguf_init_from_reader(const struct gguf_reader & gr
gguf_free(ctx);
return nullptr;
}
size_t padded_size = GGML_PAD(ggml_nbytes(&ti.t), ctx->alignment);
const size_t nbytes = ggml_nbytes(&ti.t);
// GGML_PAD adds (alignment - 1) to nbytes. when nbytes is near
// SIZE_MAX this wraps to 0 and makes the check below dead.
if (nbytes > SIZE_MAX - (ctx->alignment - 1)) {
GGML_LOG_ERROR("%s: tensor '%s' size %zu overflows after padding (alignment %zu)\n",
__func__, ti.t.name, nbytes, ctx->alignment);
gguf_free(ctx);
return nullptr;
}
const size_t padded_size = GGML_PAD(nbytes, ctx->alignment);
if (SIZE_MAX - ctx->size < padded_size) {
GGML_LOG_ERROR("%s: tensor '%s' size overflow, cannot accumulate size %zu + %zu\n",
__func__, ti.t.name, ctx->size, padded_size);
tests/test-gguf.cpp
+10
View File
@@ -40,6 +40,7 @@ enum handcrafted_file_type {
HANDCRAFTED_TENSORS_ZERO_DIM = 35 + offset_has_tensors,
HANDCRAFTED_TENSORS_NE_TOO_BIG = 40 + offset_has_tensors,
HANDCRAFTED_TENSORS_NBYTES_TOO_BIG = 45 + offset_has_tensors,
HANDCRAFTED_TENSORS_NBYTES_PAD_WRAP = 46 + offset_has_tensors,
HANDCRAFTED_TENSORS_BAD_TYPE = 50 + offset_has_tensors,
HANDCRAFTED_TENSORS_BAD_OFFSET = 60 + offset_has_tensors,
HANDCRAFTED_TENSORS_DUPLICATE_NAME = 70 + offset_has_tensors,
@@ -80,6 +81,7 @@ static std::string handcrafted_file_type_name(const enum handcrafted_file_type h
case HANDCRAFTED_TENSORS_ZERO_DIM: return "TENSORS_ZERO_DIM";
case HANDCRAFTED_TENSORS_NE_TOO_BIG: return "TENSORS_NE_TOO_BIG";
case HANDCRAFTED_TENSORS_NBYTES_TOO_BIG: return "TENSORS_NBYTES_TOO_BIG";
case HANDCRAFTED_TENSORS_NBYTES_PAD_WRAP: return "TENSORS_NBYTES_PAD_WRAP";
case HANDCRAFTED_TENSORS_BAD_TYPE: return "TENSORS_BAD_TYPE";
case HANDCRAFTED_TENSORS_BAD_OFFSET: return "TENSORS_BAD_OFFSET";
case HANDCRAFTED_TENSORS_DUPLICATE_NAME: return "TENSORS_DUPLICATE_NAME";
@@ -250,6 +252,13 @@ static FILE * get_handcrafted_file(const unsigned int seed, const enum handcraft
tensor_configs[1] = { GGML_TYPE_I8, { 0x7FFFFFFFFFFFFFC0, 1, 1, 1 } };
}
if (hft == HANDCRAFTED_TENSORS_NBYTES_PAD_WRAP) {
tensor_configs.resize(1);
// F32 with ne = [4, 2^30-1, 2^30+1, 1] so ggml_nbytes = 2^64 - 16.
// this hits the GGML_PAD wrap window: pad wraps to 0.
tensor_configs[0] = { GGML_TYPE_F32, { 4, INT64_C(1073741823), INT64_C(1073741825), 1 } };
}
if (hft == HANDCRAFTED_HEADER_BAD_N_TENSORS) {
const uint64_t n_tensors = -1;
helper_write(file, n_tensors);
@@ -774,6 +783,7 @@ static std::pair<int, int> test_handcrafted_file(const unsigned int seed) {
HANDCRAFTED_TENSORS_ZERO_DIM,
HANDCRAFTED_TENSORS_NE_TOO_BIG,
HANDCRAFTED_TENSORS_NBYTES_TOO_BIG,
HANDCRAFTED_TENSORS_NBYTES_PAD_WRAP,
HANDCRAFTED_TENSORS_BAD_TYPE,
HANDCRAFTED_TENSORS_BAD_OFFSET,
HANDCRAFTED_TENSORS_DUPLICATE_NAME,