segfault in sqlite3 (3.35.4)
(1) By tsiming on 2021-04-07 07:12:25 [link] [source]
Hi, when using sqlite3 binary built from sqlite-3.35.4 on Ubuntu 16.04, a segmentation fault error occured. There may exist some problem in the source code. poc file content: .ar -aa running command: /home/Projects/Programs/sqlite-autoconf-3350400/build/sqlite3 < poc and I got following errors: valgrind: ==1687082== Memcheck, a memory error detector ==1687082== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al. ==1687082== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info ==1687082== Command: /home/Projects/Programs/sqlite-autoconf-3350400/build/sqlite3 ==1687082== ==1687082== Invalid read of size 8 ==1687082== at 0x466417: apndOpen (shell.c:4161) ==1687082== by 0x799B43: sqlite3OsOpen (sqlite3.c:23470) ==1687082== by 0x799B43: sqlite3PagerOpen (sqlite3.c:57016) ==1687082== by 0x799B43: sqlite3BtreeOpen (sqlite3.c:67464) ==1687082== by 0x81C0E7: openDatabase (sqlite3.c:167352) ==1687082== by 0x4B8229: arDotCommand (shell.c:16797) ==1687082== by 0x4BF2B8: do_meta_command (shell.c:17607) ==1687082== by 0x4E5462: process_input (shell.c:20682) ==1687082== by 0x420EF5: main (shell.c:21524) ==1687082== Address 0x8 is not stack'd, malloc'd or (recently) free'd ==1687082== ==1687082== ==1687082== Process terminating with default action of signal 11 (SIGSEGV): dumping core ==1687082== Access not within mapped region at address 0x8 ==1687082== at 0x466417: apndOpen (shell.c:4161) ==1687082== by 0x799B43: sqlite3OsOpen (sqlite3.c:23470) ==1687082== by 0x799B43: sqlite3PagerOpen (sqlite3.c:57016) ==1687082== by 0x799B43: sqlite3BtreeOpen (sqlite3.c:67464) ==1687082== by 0x81C0E7: openDatabase (sqlite3.c:167352) ==1687082== by 0x4B8229: arDotCommand (shell.c:16797) ==1687082== by 0x4BF2B8: do_meta_command (shell.c:17607) ==1687082== by 0x4E5462: process_input (shell.c:20682) ==1687082== by 0x420EF5: main (shell.c:21524) ==1687082== If you believe this happened as a result of a stack ==1687082== overflow in your program's main thread (unlikely but ==1687082== possible), you can try to increase the size of the ==1687082== main thread stack using the --main-stacksize= flag. ==1687082== The main thread stack size used in this run was 8388608. ==1687082== ==1687082== HEAP SUMMARY: ==1687082== in use at exit: 72,650 bytes in 235 blocks ==1687082== total heap usage: 317 allocs, 82 frees, 94,943 bytes allocated ==1687082== ==1687082== LEAK SUMMARY: ==1687082== definitely lost: 0 bytes in 0 blocks ==1687082== indirectly lost: 0 bytes in 0 blocks ==1687082== possibly lost: 24 bytes in 1 blocks ==1687082== still reachable: 72,626 bytes in 234 blocks ==1687082== of which reachable via heuristic: ==1687082== length64 : 72,520 bytes in 232 blocks ==1687082== suppressed: 0 bytes in 0 blocks ==1687082== Rerun with --leak-check=full to see details of leaked memory ==1687082== ==1687082== For counts of detected and suppressed errors, rerun with: -v ==1687082== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0) Segmentation fault gdb: Starting program: /home/Projects/Programs/sqlite-autoconf-3350400/build/sqlite3 < poc [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". Program received signal SIGSEGV, Segmentation fault. 0x0000000000466417 in apndOpen (pApndVfs=<optimized out>, zName=<optimized out>, pFile=0xcd6090, flags=257, pOutFlags=<optimized out>) at ../shell.c:4161 4161 pBaseFile->pMethods->xClose(pBaseFile); asan: ASAN:SIGSEGV ================================================================= ==3676881==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000008 (pc 0x00000040fe9a bp 0x7fffdffd1a90 sp 0x7fffdffd19c0 T0) #0 0x40fe99 in apndOpen (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x40fe99) #1 0x46bfde in sqlite3OsOpen (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x46bfde) #2 0x4a33c7 in sqlite3PagerOpen (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x4a33c7) #3 0x4c1016 in sqlite3BtreeOpen (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x4c1016) #4 0x652d3e in openDatabase (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x652d3e) #5 0x65352c in sqlite3_open_v2 (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x65352c) #6 0x44cd00 in arDotCommand (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x44cd00) #7 0x450e75 in do_meta_command (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x450e75) #8 0x46116f in process_input (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x46116f) #9 0x463d37 in main (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x463d37) #10 0x7f75c916883f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2083f) #11 0x403798 in _start (/home/Projects/Programs/sqlite-autoconf-3350400/build-asan/sqlite3+0x403798) AddressSanitizer can not provide additional info. SUMMARY: AddressSanitizer: SEGV ??:0 apndOpen ==3676881==ABORTING
(2) By Larry Brasfield (larrybr) on 2021-04-07 09:46:22 in reply to 1 [link] [source]
Thanks for the bug report. This problem has been fixed.
Until you get another release build, version preview, or the above fix from the repository, you can work around this problem by attempting to use the AppendVFS only on already existing, readable files. For example, instead of writing
.ar -a NowhereToBeFound
as a shell meta-command, write
.ar -a ReadableFileToBeAppended
. The next shell version will handle the former more gracefully.
(3) By anonymous on 2021-04-07 13:26:20 in reply to 2 [link] [source]
Will this result in 3.35.5, or will you wait until 3.36?
(4) By anonymous on 2021-04-07 13:30:03 in reply to 2 [link] [source]
Hi Larry. SQLite fixes typically come with tests, to keep that 100% line and branch coverage trademark testing.
ext/ code come with different guarantees? Or are those tests part of the non-public test suite?
Not complaining mind you. Just wondering. Thanks.
(5) By Larry Brasfield (larrybr) on 2021-04-07 14:22:32 in reply to 4 [source]
There is no intention or attitude, that I am aware of, that those parts of the project are to be second class with respect to quality. Testing is another matter.
In this case, (considering further tests for ext/misc/appendvfs.c), I elected to think further upon what was appropriate rather than just writing more tests. The test suite takes hours to run on each platform, even on modern hardware, and it is not clear to me that the affected shim merits the same degree of coverage that the SQLite library itself gets. This is not to suggest it will get no more tests; just that what more it should get is a more nuanced issue than a seg-fault fix.
(6) By Richard Hipp (drh) on 2021-04-07 14:31:37 in reply to 4 [link] [source]
The 100% MC/DC guarantee only applies to the SQLite core. The extensions are not tested to the same level of rigor.