##// END OF EJS Templates
revlog: add a mechanism to verify expected file position before appending...
revlog: add a mechanism to verify expected file position before appending If someone uses `hg debuglocks`, or some non-hg process writes to the .hg directory without respecting the locks, or if the repo's on a networked filesystem, it's possible for the revlog code to write out corrupted data. The form of this corruption can vary depending on what data was written and how that happened. We are in the "networked filesystem" case (though I've had users also do this to themselves with the "`hg debuglocks`" scenario), and most often see this with the changelog. What ends up happening is we produce two items (let's call them rev1 and rev2) in the .i file that have the same linkrev, baserev, and offset into the .d file, while the data in the .d file is appended properly. rev2's compressed_size is accurate for rev2, but when we go to decompress the data in the .d file, we use the offset that's recorded in the index file, which is the same as rev1, and attempt to decompress rev2.compressed_size bytes of rev1's data. This usually does not succeed. :) When using inline data, this also fails, though I haven't investigated why too closely. This shows up as a "patch decode" error. I believe what's happening there is that we're basically ignoring the offset field, getting the data properly, but since baserev != rev, it thinks this is a delta based on rev (instead of a full text) and can't actually apply it as such. For now, I'm going to make this an optional component and default it to entirely off. I may increase the default severity of this in the future, once I've enabled it for my users and we gain more experience with it. Luckily, most of my users have a versioned filesystem and can roll back to before the corruption has been written, it's just a hassle to do so and not everyone knows how (so it's a support burden). Users on other filesystems will not have that luxury, and this can cause them to have a corrupted repository that they are unlikely to know how to resolve, and they'll see this as a data-loss event. Refusing to create the corruption is a much better user experience. This mechanism is not perfect. There may be false-negatives (racy writes that are not detected). There should not be any false-positives (non-racy writes that are detected as such). This is not a mechanism that makes putting a repo on a networked filesystem "safe" or "supported", just *less* likely to cause corruption. Differential Revision: https://phab.mercurial-scm.org/D9952

File last commit:

r44546:dc9b5348 default
r47349:e9901d01 default
Show More
cext.c
212 lines | 6.6 KiB | text/x-c | CLexer
#define PY_SSIZE_T_CLEAN
#include <Python.h>
#include "lib/sha1.h"
#if PY_MAJOR_VERSION >= 3
#define IS_PY3K
#endif
/* helper to switch things like string literal depending on Python version */
#ifdef IS_PY3K
#define PY23(py2, py3) py3
#else
#define PY23(py2, py3) py2
#endif
static char sha1dc_doc[] = "Efficient detection of SHA1 collision constructs.";
/* clang-format off */
typedef struct {
PyObject_HEAD
SHA1_CTX ctx;
} pysha1ctx;
/* clang-format on */
static int pysha1ctx_init(pysha1ctx *self, PyObject *args)
{
Py_buffer data;
data.obj = NULL;
SHA1DCInit(&(self->ctx));
/* We don't want "safe" sha1s, wherein sha1dc can give you a
different hash for something that's trying to give you a
collision. We just want to detect collisions.
*/
SHA1DCSetSafeHash(&(self->ctx), 0);
if (!PyArg_ParseTuple(args, PY23("|s*", "|y*"), &data)) {
return -1;
}
if (data.obj) {
if (!PyBuffer_IsContiguous(&data, 'C') || data.ndim > 1) {
PyErr_SetString(PyExc_BufferError,
"buffer must be contiguous and single dimension");
PyBuffer_Release(&data);
return -1;
}
SHA1DCUpdate(&(self->ctx), data.buf, data.len);
PyBuffer_Release(&data);
}
return 0;
}
static void pysha1ctx_dealloc(pysha1ctx *self)
{
PyObject_Del(self);
}
static PyObject *pysha1ctx_update(pysha1ctx *self, PyObject *args)
{
Py_buffer data;
if (!PyArg_ParseTuple(args, PY23("s*", "y*"), &data)) {
return NULL;
}
if (!PyBuffer_IsContiguous(&data, 'C') || data.ndim > 1) {
PyErr_SetString(PyExc_BufferError,
"buffer must be contiguous and single dimension");
PyBuffer_Release(&data);
return NULL;
}
SHA1DCUpdate(&(self->ctx), data.buf, data.len);
PyBuffer_Release(&data);
Py_RETURN_NONE;
}
/* it is intentional that this take a ctx by value, as that clones the
context so we can keep using .update() without poisoning the state
with padding.
*/
static int finalize(SHA1_CTX ctx, unsigned char *hash_out)
{
if (SHA1DCFinal(hash_out, &ctx)) {
PyErr_SetString(PyExc_OverflowError,
"sha1 collision attack detected");
return 0;
}
return 1;
}
static PyObject *pysha1ctx_digest(pysha1ctx *self)
{
unsigned char hash[20];
if (!finalize(self->ctx, hash)) {
return NULL;
}
return PyBytes_FromStringAndSize((char *)hash, 20);
}
static PyObject *pysha1ctx_hexdigest(pysha1ctx *self)
{
static const char hexdigit[] = "0123456789abcdef";
unsigned char hash[20];
char hexhash[40];
int i;
if (!finalize(self->ctx, hash)) {
return NULL;
}
for (i = 0; i < 20; ++i) {
hexhash[i * 2] = hexdigit[hash[i] >> 4];
hexhash[i * 2 + 1] = hexdigit[hash[i] & 15];
}
return PY23(PyString_FromStringAndSize, PyUnicode_FromStringAndSize)(hexhash, 40);
}
static PyTypeObject sha1ctxType;
static PyObject *pysha1ctx_copy(pysha1ctx *self)
{
pysha1ctx *clone = (pysha1ctx *)PyObject_New(pysha1ctx, &sha1ctxType);
if (!clone) {
return NULL;
}
clone->ctx = self->ctx;
return (PyObject *)clone;
}
static PyMethodDef pysha1ctx_methods[] = {
{"update", (PyCFunction)pysha1ctx_update, METH_VARARGS,
"Update this hash object's state with the provided bytes."},
{"digest", (PyCFunction)pysha1ctx_digest, METH_NOARGS,
"Return the digest value as a string of binary data."},
{"hexdigest", (PyCFunction)pysha1ctx_hexdigest, METH_NOARGS,
"Return the digest value as a string of hexadecimal digits."},
{"copy", (PyCFunction)pysha1ctx_copy, METH_NOARGS,
"Return a copy of the hash object."},
{NULL},
};
/* clang-format off */
static PyTypeObject sha1ctxType = {
PyVarObject_HEAD_INIT(NULL, 0) /* header */
"sha1dc.sha1", /* tp_name */
sizeof(pysha1ctx), /* tp_basicsize */
0, /* tp_itemsize */
(destructor)pysha1ctx_dealloc, /* tp_dealloc */
0, /* tp_print */
0, /* tp_getattr */
0, /* tp_setattr */
0, /* tp_compare */
0, /* tp_repr */
0, /* tp_as_number */
0, /* tp_as_sequence */
0, /* tp_as_mapping */
0, /* tp_hash */
0, /* tp_call */
0, /* tp_str */
0, /* tp_getattro */
0, /* tp_setattro */
0, /* tp_as_buffer */
Py_TPFLAGS_DEFAULT, /* tp_flags */
"sha1 implementation that looks for collisions", /* tp_doc */
0, /* tp_traverse */
0, /* tp_clear */
0, /* tp_richcompare */
0, /* tp_weaklistoffset */
0, /* tp_iter */
0, /* tp_iternext */
pysha1ctx_methods, /* tp_methods */
0, /* tp_members */
0, /* tp_getset */
0, /* tp_base */
0, /* tp_dict */
0, /* tp_descr_get */
0, /* tp_descr_set */
0, /* tp_dictoffset */
(initproc)pysha1ctx_init, /* tp_init */
0, /* tp_alloc */
};
/* clang-format on */
static PyMethodDef methods[] = {
{NULL, NULL},
};
static void module_init(PyObject *mod)
{
sha1ctxType.tp_new = PyType_GenericNew;
if (PyType_Ready(&sha1ctxType) < 0) {
return;
}
Py_INCREF(&sha1ctxType);
PyModule_AddObject(mod, "sha1", (PyObject *)&sha1ctxType);
}
#ifdef IS_PY3K
static struct PyModuleDef sha1dc_module = {PyModuleDef_HEAD_INIT, "sha1dc",
sha1dc_doc, -1, methods};
PyMODINIT_FUNC PyInit_sha1dc(void)
{
PyObject *mod = PyModule_Create(&sha1dc_module);
module_init(mod);
return mod;
}
#else
PyMODINIT_FUNC initsha1dc(void)
{
PyObject *mod = Py_InitModule3("sha1dc", methods, sha1dc_doc);
module_init(mod);
}
#endif