From: Safin Timur via Tarantool-patches <tarantool-patches@dev.tarantool.org> To: Mergen Imeev <imeevma@tarantool.org> Cc: tarantool-patches@dev.tarantool.org Subject: Re: [Tarantool-patches] [PATCH v1 1/1] sql: fix a segfault in hex() on receiving zeroblob Date: Fri, 3 Sep 2021 22:19:56 +0300 [thread overview] Message-ID: <73f26e5c-9374-682a-5787-0da49b32953c@tarantool.org> (raw) In-Reply-To: <20210901084450.GA111802@tarantool.org> On 01.09.2021 11:44, Mergen Imeev wrote: > Hi! Thank you for the review. My answers below. > > On Tue, Aug 31, 2021 at 10:32:46PM +0300, Timur Safin wrote: >> I may miss something obvious, but prior version of a code >> with pBlob and n was much shorter, compacter and more readable. >> I'm curious, why do you prefer to always use argv[0]->n and >> argv[0]->z instead? >> > If we talk about the old function, then it really looks simpler. However, it did > not work correctly and also made some unnecessary changes to the arguments. You > can compare to the fixed version of old function on this branch: > imeevma/gh-6113-fix-hex-segfault-2.8 (which I also sent you for review). You will > see much less difference there. I meant that newer code was a little bit .. mouthful, with unnecessary code substitution and visual noise which harmed readability. Here is an example of version which is not using argv[0]->.. wherever we refer to fields. ---------------------------------------------------- /** Implementation of the HEX() SQL built-in function. */ static void func_hex(struct sql_context *ctx, int argc, struct Mem **argv) { assert(argc == 1); (void)argc; if (argv[0]->type == MEM_TYPE_NULL) return mem_set_null(ctx->pOut); int n = argv[0]->n; int zero_len = argv[0]->u.nZero; assert(argv[0]->type == MEM_TYPE_BIN && n >= 0); assert((argv[0]->flags & MEM_Zero) == 0 || zero_len >= 0); uint32_t size = 2 * n; if ((argv[0]->flags & MEM_Zero) != 0) size += 2 * zero_len; if (size == 0) return mem_set_str0_static(ctx->pOut, ""); char *str = sqlDbMallocRawNN(sql_get(), size); if (str == NULL) { ctx->is_aborted = true; return; } for (int i = 0; i < n; ++i) { char c = argv[0]->z[i]; str[2 * i] = hexdigits[(c >> 4) & 0xf]; str[2 * i + 1] = hexdigits[c & 0xf]; } if ((argv[0]->flags & MEM_Zero) != 0) memset(&str[2 * n], '0', 2 * zero_len); mem_set_str_allocated(ctx->pOut, str, size); } ---------------------------------------------------- It's more resembling original code (and that was done intentionally). Also (and I didn't change it in the sample) there is apparent missing check for SQL_LIMIT_LENGTH limit which used to be done in contextMalloc() before, but now is missing once we use sqlDbMallocRawNN(). I assume we better return this check (once again as a proper wrapper which contextMalloc() essentially was). > >> Also, it seems to me we better to limit the number of bytes customer >> may request to allocate from HEX()? What about to check against SQL_LIMIT_LENGTH? >> > This check is performed in the OP_BuiltinFunction opcode. That's nice, so it's not a problem then. > >> Thanks, >> Timur >> >>> -----Original Message----- >>> From: imeevma@tarantool.org <imeevma@tarantool.org> >>> Sent: Monday, August 30, 2021 9:31 AM >>> To: tsafin@tarantool.org >>> Cc: tarantool-patches@dev.tarantool.org >>> Subject: [PATCH v1 1/1] sql: fix a segfault in hex() on receiving >>> zeroblob >>> >>> This patch fixes a segmentation fault when zeroblob is received by >>> the >>> SQL built-in HEX() function. >>> >>> Closes #6113 >>> --- >>> https://github.com/tarantool/tarantool/issues/6113 >>> https://github.com/tarantool/tarantool/tree/imeevma/gh-6113-fix-hex- >>> segfault-2.10 >>> >>> .../gh-6113-fix-segfault-in-hex-func.md | 5 ++ >>> src/box/sql/func.c | 75 ++++++++++------- >>> -- >>> test/sql-tap/engine.cfg | 1 + >>> ...gh-6113-assert-in-hex-on-zeroblob.test.lua | 13 ++++ >>> 4 files changed, 58 insertions(+), 36 deletions(-) >>> create mode 100644 changelogs/unreleased/gh-6113-fix-segfault-in- >>> hex-func.md >>> create mode 100755 test/sql-tap/gh-6113-assert-in-hex-on- >>> zeroblob.test.lua >>> >>> diff --git a/changelogs/unreleased/gh-6113-fix-segfault-in-hex- >>> func.md b/changelogs/unreleased/gh-6113-fix-segfault-in-hex-func.md >>> new file mode 100644 >>> index 000000000..c59be4d96 >>> --- /dev/null >>> +++ b/changelogs/unreleased/gh-6113-fix-segfault-in-hex-func.md >>> @@ -0,0 +1,5 @@ >>> +## bugfix/sql >>> + >>> +* The HEX() SQL built-in function now does not throw an assert on >>> receiving >>> + varbinary values that consist of zero-bytes (gh-6113). >>> + >>> diff --git a/src/box/sql/func.c b/src/box/sql/func.c >>> index c063552d6..fa2a2c245 100644 >>> --- a/src/box/sql/func.c >>> +++ b/src/box/sql/func.c >>> @@ -53,6 +53,44 @@ >>> static struct mh_strnptr_t *built_in_functions = NULL; >>> static struct func_sql_builtin **functions; >>> >>> +/** Array for converting from half-bytes into ASCII hex digits. */ >>> +static const char hexdigits[] = { >>> + '0', '1', '2', '3', '4', '5', '6', '7', >>> + '8', '9', 'A', 'B', 'C', 'D', 'E', 'F' >>> +}; >>> + >>> +/** Implementation of the HEX() SQL built-in function. */ >>> +static void >>> +func_hex(struct sql_context *ctx, int argc, struct Mem **argv) >>> +{ >>> + assert(argc == 1); >>> + (void)argc; >>> + if (argv[0]->type == MEM_TYPE_NULL) >>> + return mem_set_null(ctx->pOut); >>> + >>> + assert(argv[0]->type == MEM_TYPE_BIN && argv[0]->n >= 0); >>> + assert((argv[0]->flags & MEM_Zero) == 0 || argv[0]->u.nZero >= >>> 0); >>> + uint32_t size = 2 * argv[0]->n; >>> + if ((argv[0]->flags & MEM_Zero) != 0) >>> + size += 2 * argv[0]->u.nZero; >>> + if (size == 0) >>> + return mem_set_str0_static(ctx->pOut, ""); >>> + >>> + char *str = sqlDbMallocRawNN(sql_get(), size); >>> + if (str == NULL) { >>> + ctx->is_aborted = true; >>> + return; >>> + } >>> + for (int i = 0; i < argv[0]->n; ++i) { >>> + char c = argv[0]->z[i]; >>> + str[2 * i] = hexdigits[(c >> 4) & 0xf]; >>> + str[2 * i + 1] = hexdigits[c & 0xf]; >>> + } >>> + if ((argv[0]->flags & MEM_Zero) != 0) >>> + memset(&str[2 * argv[0]->n], '0', 2 * argv[0]->u.nZero); >>> + mem_set_str_allocated(ctx->pOut, str, size); >>> +} >>> + >>> static const unsigned char * >>> mem_as_ustr(struct Mem *mem) >>> { >>> @@ -1072,14 +1110,6 @@ sql_func_version(struct sql_context *context, >>> sql_result_text(context, tarantool_version(), -1, SQL_STATIC); >>> } >>> >>> -/* Array for converting from half-bytes (nybbles) into ASCII hex >>> - * digits. >>> - */ >>> -static const char hexdigits[] = { >>> - '0', '1', '2', '3', '4', '5', '6', '7', >>> - '8', '9', 'A', 'B', 'C', 'D', 'E', 'F' >>> -}; >>> - >>> /* >>> * Implementation of the QUOTE() function. This function takes a >>> single >>> * argument. If the argument is numeric, the return value is the >>> same as >>> @@ -1233,33 +1263,6 @@ charFunc(sql_context * context, int argc, >>> sql_value ** argv) >>> sql_result_text64(context, (char *)z, zOut - z, sql_free); >>> } >>> >>> -/* >>> - * The hex() function. Interpret the argument as a blob. Return >>> - * a hexadecimal rendering as text. >>> - */ >>> -static void >>> -hexFunc(sql_context * context, int argc, sql_value ** argv) >>> -{ >>> - int i, n; >>> - const unsigned char *pBlob; >>> - char *zHex, *z; >>> - assert(argc == 1); >>> - UNUSED_PARAMETER(argc); >>> - pBlob = mem_as_bin(argv[0]); >>> - n = mem_len_unsafe(argv[0]); >>> - assert(pBlob == mem_as_bin(argv[0])); /* No encoding change */ >>> - z = zHex = contextMalloc(context, ((i64) n) * 2 + 1); >>> - if (zHex) { >>> - for (i = 0; i < n; i++, pBlob++) { >>> - unsigned char c = *pBlob; >>> - *(z++) = hexdigits[(c >> 4) & 0xf]; >>> - *(z++) = hexdigits[c & 0xf]; >>> - } >>> - *z = 0; >>> - sql_result_text(context, zHex, n * 2, sql_free); >>> - } >>> -} >>> - >>> /* >>> * The zeroblob(N) function returns a zero-filled blob of size N >>> bytes. >>> */ >>> @@ -2034,7 +2037,7 @@ static struct sql_func_definition definitions[] >>> = { >>> {"GROUP_CONCAT", 2, {FIELD_TYPE_VARBINARY, >>> FIELD_TYPE_VARBINARY}, >>> FIELD_TYPE_VARBINARY, groupConcatStep, groupConcatFinalize}, >>> >>> - {"HEX", 1, {FIELD_TYPE_VARBINARY}, FIELD_TYPE_STRING, hexFunc, >>> NULL}, >>> + {"HEX", 1, {FIELD_TYPE_VARBINARY}, FIELD_TYPE_STRING, func_hex, >>> NULL}, >>> {"IFNULL", 2, {FIELD_TYPE_ANY, FIELD_TYPE_ANY}, >>> FIELD_TYPE_SCALAR, >>> sql_builtin_stub, NULL}, >>> Regards, Timur
next prev parent reply other threads:[~2021-09-03 19:20 UTC|newest] Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-08-30 6:30 Mergen Imeev via Tarantool-patches 2021-08-31 19:32 ` Timur Safin via Tarantool-patches 2021-09-01 8:44 ` Mergen Imeev via Tarantool-patches 2021-09-03 19:19 ` Safin Timur via Tarantool-patches [this message] 2021-09-06 9:45 ` Mergen Imeev via Tarantool-patches 2021-09-06 20:32 ` Safin Timur via Tarantool-patches 2021-09-07 9:16 ` Mergen Imeev via Tarantool-patches -- strict thread matches above, loose matches on Subject: below -- 2021-10-05 12:49 Mergen Imeev via Tarantool-patches 2021-08-30 6:20 Mergen Imeev via Tarantool-patches 2021-09-03 19:20 ` Safin Timur via Tarantool-patches 2021-08-26 11:11 Mergen Imeev via Tarantool-patches 2021-08-26 20:42 ` Vladislav Shpilevoy via Tarantool-patches 2021-08-27 8:26 ` Mergen Imeev via Tarantool-patches 2021-08-27 21:31 ` Vladislav Shpilevoy via Tarantool-patches 2021-08-26 11:10 Mergen Imeev via Tarantool-patches 2021-08-26 20:31 ` Vladislav Shpilevoy via Tarantool-patches 2021-08-27 7:54 ` Mergen Imeev via Tarantool-patches 2021-08-27 21:52 ` Vladislav Shpilevoy via Tarantool-patches
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=73f26e5c-9374-682a-5787-0da49b32953c@tarantool.org \ --to=tarantool-patches@dev.tarantool.org \ --cc=imeevma@tarantool.org \ --cc=tsafin@tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH v1 1/1] sql: fix a segfault in hex() on receiving zeroblob' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox