View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0008787 | GNUnet | namestore service | public | 2024-05-03 03:23 | 2024-06-08 12:03 |
Reporter | ulfvonbelow | Assigned To | schanzen | ||
Priority | normal | Severity | major | Reproducibility | always |
Status | closed | Resolution | fixed | ||
Platform | x86-64 | OS | Guix System | OS Version | a1d711c92e |
Product Version | 0.21.1 | ||||
Fixed in Version | 0.21.2 | ||||
Summary | 0008787: use-after-free in handle_edit_record_set | ||||
Description | The editor_hint string that is passed in to GNUNET_NAMESTORE_RecordIterators isn't guaranteed to stay live between when the iterator is called and when GNUNET_NAMESTORE_PluginFunctions.edit_records returns. In fact, currently, it is guaranteed to have already been freed in the sqlite backend. | ||||
Steps To Reproduce | 1. ./configure --enable-sanitizer ; make ; make check 2. Observe use-after-free in gnunet-service-namestore. | ||||
Additional Information | The already-freed editor hint string is copied up to the first null byte and sent to a client. I haven't looked into how deep the security implications of this go, so I'm setting this to private for now. While observing this bug, the crash in gnunet-service-namestore caused by the sanitizer also caused a use-after-free to be observed in test_namestore_api_edit_records. This is because no error seems to have been reported despite nsqe already having been freed in force_reconnect in src/service/namestore/namestore_api.c. In fact, there seems to be no ability to specify an error callback with GNUNET_NAMESTORE_record_set_edit_begin, although the documentation next to the prototype in src/include/gnunet_namestore_service.h mentions both error_cb and error_cb_cls. Anyway, the endbadly timeout in the test eventually hits and it attempts to cancel nsqe, which has already been freed. I mention this here because it's not exactly a bug so much as a limitation. After fixing this using the attached patch, it was discovered that editor_hint was actually never properly implemented in plugin_namestore_sqlite, so fixing this will actually cause test_namestore_api_edit_records to get the actual editor hint string and fail because of it. Will report that separately also. | ||||
Tags | No tags attached. | ||||
Attached Files | 0001-service-namestore-fix-use-after-free-in-handle_edit_.patch (2,739 bytes)
From 0c328d4b3a8c2c005f3798b1048b0862d1cd34bd Mon Sep 17 00:00:00 2001 From: ulfvonbelow <striness@tilde.club> Date: Thu, 2 May 2024 16:05:00 -0500 Subject: [PATCH] service: namestore: fix use-after-free in handle_edit_record_set. The editor_hint string that is passed in to GNUNET_NAMESTORE_RecordIterators isn't guaranteed to stay live between when the iterator is called and when GNUNET_NAMESTORE_PluginFunctions.edit_records returns. So lookup_it should strdup that string, and anything that uses lookup_it should manage its lifetime. Currently that's just handle_edit_record_set and handle_record_lookup. --- src/service/namestore/gnunet-service-namestore.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/src/service/namestore/gnunet-service-namestore.c b/src/service/namestore/gnunet-service-namestore.c index f29058d0e..448a9d2b3 100644 --- a/src/service/namestore/gnunet-service-namestore.c +++ b/src/service/namestore/gnunet-service-namestore.c @@ -1157,7 +1157,8 @@ lookup_it (void *cls, if (0 != strcmp (label, rlc->label)) return; rlc->found = GNUNET_YES; - rlc->editor_hint = editor_hint; + if (NULL == rlc->editor_hint) + rlc->editor_hint = GNUNET_strdup (editor_hint); if (GNUNET_OK != GNUNET_GNSRECORD_normalize_record_set (rlc->label, rd_nf, rd_count_nf, @@ -1332,6 +1333,7 @@ handle_edit_record_set (void *cls, const struct EditRecordSetMessage *er_msg) return; } name_len = strlen (conv_name) + 1; + rlc.editor_hint = NULL; rlc.label = conv_name; GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Looking up without filter\n"); @@ -1370,6 +1372,7 @@ handle_edit_record_set (void *cls, const struct EditRecordSetMessage *er_msg) GNUNET_memcpy ((char*) &rer_msg[1] + old_editor_hint_len, rlc.res_rd, rlc.rd_ser_len); GNUNET_MQ_send (nc->mq, env); + GNUNET_free (rlc.editor_hint); GNUNET_free (rlc.res_rd); GNUNET_free (conv_name); } @@ -1566,6 +1569,7 @@ handle_record_lookup (void *cls, const struct LabelLookupMessage *ll_msg) return; } name_len = strlen (conv_name) + 1; + rlc.editor_hint = NULL; rlc.label = conv_name; GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Looking up with filter %u\n", ntohs (ll_msg->filter)); @@ -1601,6 +1605,7 @@ handle_record_lookup (void *cls, const struct LabelLookupMessage *ll_msg) GNUNET_memcpy (res_name, conv_name, name_len); GNUNET_memcpy (&res_name[name_len], rlc.res_rd, rlc.rd_ser_len); GNUNET_MQ_send (nc->mq, env); + GNUNET_free (rlc.editor_hint); GNUNET_free (rlc.res_rd); GNUNET_free (conv_name); } -- 2.41.0 | ||||
Date Modified | Username | Field | Change |
---|---|---|---|
2024-05-03 03:23 | ulfvonbelow | New Issue | |
2024-05-03 03:23 | ulfvonbelow | File Added: 0001-service-namestore-fix-use-after-free-in-handle_edit_.patch | |
2024-05-05 13:58 | schanzen | Assigned To | => schanzen |
2024-05-05 13:58 | schanzen | Status | new => resolved |
2024-05-05 13:58 | schanzen | Resolution | open => fixed |
2024-05-05 13:58 | schanzen | Fixed in Version | => 0.21.2 |
2024-05-05 13:58 | schanzen | Note Added: 0022341 | |
2024-05-05 13:58 | schanzen | View Status | private => public |
2024-06-08 12:03 | schanzen | Note Added: 0022558 | |
2024-06-08 12:03 | schanzen | Status | resolved => closed |