From patchwork Thu Oct 25 21:32:58 2012 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Dodji Seketeli X-Patchwork-Id: 194313 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) by ozlabs.org (Postfix) with SMTP id 3F9B52C009A for ; Fri, 26 Oct 2012 08:33:18 +1100 (EST) Comment: DKIM? See http://www.dkim.org DKIM-Signature: v=1; a=rsa-sha1; c=relaxed/relaxed; d=gcc.gnu.org; s=default; x=1351805599; h=Comment: DomainKey-Signature:Received:Received:Received:Received:Received: Received:From:To:Cc:Subject:References:Date:In-Reply-To: Message-ID:User-Agent:MIME-Version:Content-Type:Mailing-List: Precedence:List-Id:List-Unsubscribe:List-Archive:List-Post: List-Help:Sender:Delivered-To; bh=gVSGmsVIQP99BeDgNS+cX6YJOsk=; b=HvTlztzGBA22NkbBsgCTdD809hUGC/hAKIxYvB++C9XN+vuAGDsDf41r2F2VHh pKCOk4WnaXqSfpwFNyAPHUBQmsw+7zndBAOJMF1mmjcFCA4S6o/lGhptl1Gcr0tc HTk1oVvZxs/R7MYTVH9AYuhVQYw+ohm1gqTDQUuAGhUv4= Comment: DomainKeys? See http://antispam.yahoo.com/domainkeys DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=default; d=gcc.gnu.org; h=Received:Received:X-SWARE-Spam-Status:X-Spam-Check-By:Received:Received:Received:Received:From:To:Cc:Subject:References:X-URL:Date:In-Reply-To:Message-ID:User-Agent:MIME-Version:Content-Type:Mailing-List:Precedence:List-Id:List-Unsubscribe:List-Archive:List-Post:List-Help:Sender:Delivered-To; b=IDJQooc2e8ObQde9bYXMAFQNi/OVFw+edUB23gpOJQ6gGWMa0GX/UMFPmjbNjM WuPI3/igkwJOd6M9GOryZr3qxx3aFZU3S+6k9RnrYjl4aFlbGUvuePS/SKApeltf u9ep7P41+W7eZzLRXB6AM2QiNYDWz+6zM0WWaHPKQw4Fk=; Received: (qmail 22600 invoked by alias); 25 Oct 2012 21:33:14 -0000 Received: (qmail 22591 invoked by uid 22791); 25 Oct 2012 21:33:13 -0000 X-SWARE-Spam-Status: No, hits=-6.3 required=5.0 tests=AWL, BAYES_00, KHOP_RCVD_UNTRUST, RCVD_IN_DNSWL_HI, RCVD_IN_HOSTKARMA_W, RP_MATCHES_RCVD, SPF_HELO_PASS, TW_LV, TW_TM X-Spam-Check-By: sourceware.org Received: from mx1.redhat.com (HELO mx1.redhat.com) (209.132.183.28) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Thu, 25 Oct 2012 21:33:01 +0000 Received: from int-mx12.intmail.prod.int.phx2.redhat.com (int-mx12.intmail.prod.int.phx2.redhat.com [10.5.11.25]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id q9PLX1mB018206 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Thu, 25 Oct 2012 17:33:01 -0400 Received: from localhost (ovpn-116-36.ams2.redhat.com [10.36.116.36]) by int-mx12.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id q9PLWxMP027066 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 25 Oct 2012 17:33:00 -0400 Received: by localhost (Postfix, from userid 1000) id 2A03B2C0104; Thu, 25 Oct 2012 23:32:59 +0200 (CEST) From: Dodji Seketeli To: Jakub Jelinek Cc: GCC Patches , Diego Novillo , Xinliang David Li , Wei Mi Subject: Re: [PATCH 3/3] [asan] Instrument built-in memory access function calls References: <87liex8hjt.fsf@redhat.com> <8762618ha6.fsf@redhat.com> <20121023160323.GU1752@tucnak.redhat.com> <87bofr6h2c.fsf@redhat.com> <20121024161455.GI1752@tucnak.redhat.com> X-URL: http://www.redhat.com Date: Thu, 25 Oct 2012 23:32:58 +0200 In-Reply-To: <20121024161455.GI1752@tucnak.redhat.com> (Jakub Jelinek's message of "Wed, 24 Oct 2012 18:14:55 +0200") Message-ID: <87fw52i6et.fsf@redhat.com> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.1 (gnu/linux) MIME-Version: 1.0 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Delivered-To: mailing list gcc-patches@gcc.gnu.org Jakub Jelinek writes: > case BUILT_IN_ATOMIC_ALWAYS_LOCK_FREE: > case BUILT_IN_ATOMIC_IS_LOCK_FREE: > I think don't touch the memory at all (or not necessarily), > and IMHO you don't want to handle the BUILT_IN_*_N variants either, > those are just FE builtins that are lowered to the corresponding > _{1,2,4,8,16} variants. Fixed. > > > @@ -661,13 +702,477 @@ instrument_derefs (gimple_stmt_iterator *iter, tree t, > > int volatilep = 0, unsignedp = 0; > > get_inner_reference (t, &bitsize, &bitpos, &offset, > > &mode, &unsignedp, &volatilep, false); > > - if (bitpos != 0 || bitsize != size_in_bytes * BITS_PER_UNIT) > > + if (bitpos % BITS_PER_UNIT || bitsize != size_in_bytes * BITS_PER_UNIT) > > return; > > Shouldn't that be bitpos % (size_in_bytes * BITS_PER_UNIT) ? > 1 byte or 2 byte access at bitpos 80 is fine, but 4 byte access is not. Of course it should be that. I have updated the patch. Jakub Jelinek writes: > On Wed, Oct 24, 2012 at 05:16:26PM +0200, Dodji Seketeli wrote: >> Jakub Jelinek writes: >> >> >> For 'strlen', can the memory check be done at the end of the string >> >> using the returned length? >> > >> > Guess strlen is commonly expanded inline, so it would be worthwhile to check >> > the shadow memory after the call (well, we could check the first byte >> > before the call and the last one after the call). >> >> How do I get the result of the (strlen) call in gimple? > > That is gimple_call_lhs (call_stmt). > So for insturmenting strlen, you want to ammend: > tmp = strlen (ptr); > as > asan_addr_check (ptr); > tmp = strlen (ptr); > asan_addr_check (ptr + tmp); Right. I have done that in the updated patch. So, to ease the review, I am re-posting the updated patch, below with its amended introductory "cover letter" and Changelog. Thanks. This patch instruments many memory access patterns through builtins. Basically, for a call like: __builtin_memset (from, 0, n_bytes); the patch would only instrument the accesses at the beginning and at the end of the memory region [from, from + n_bytes]. This is the strategy used by the llvm implementation of asan. This instrumentation is done for all the memory access builtin functions that expose a well specified memory region -- one that explicitly states the number of bytes accessed in the region. A special treatment is used for __builtin_strlen. The patch instruments the access to the first byte of its argument, as well as the access to the byte (of the argument) at the offset returned by strlen. For the __sync_* and __atomic* calls the patch instruments the access to the bytes pointed to by the argument. While doing this, I have added a new parameter to build_check_stmt to decide whether to insert the instrumentation code before or after the statement iterator. This allows us to do away with the gsi_{next,prev} dance we were doing in the callers of this function. Tested by running cc1 -fasan on variations of simple programs like: int foo () { char foo[10] = {0}; foo[0] = 't'; foo[1] = 'e'; foo[2] = 's'; foo[3] = 't'; int l = __builtin_strlen (foo); int n = sizeof (foo); __builtin_memset (&foo[4], 0, n - 4); __sync_fetch_and_add (&foo[11], 1); return l; } and by starring at the gimple output which for this function is: ;; Function foo (foo, funcdef_no=0, decl_uid=1714, cgraph_uid=0) foo () { int n; int l; char foo[10]; int D.1723; long unsigned int D.1722; int D.1721; long unsigned int D.1720; long unsigned int _1; int _4; long unsigned int _5; int _6; char * _7; unsigned long _8; unsigned long _9; unsigned long _10; signed char * _11; signed char _12; _Bool _13; unsigned long _14; signed char _15; _Bool _16; _Bool _17; char * _18; unsigned long _19; unsigned long _20; unsigned long _21; signed char * _22; signed char _23; _Bool _24; unsigned long _25; signed char _26; _Bool _27; _Bool _28; char * _29; unsigned long _30; unsigned long _31; unsigned long _32; signed char * _33; signed char _34; _Bool _35; unsigned long _36; signed char _37; _Bool _38; _Bool _39; char * _40; unsigned long _41; unsigned long _42; unsigned long _43; signed char * _44; signed char _45; _Bool _46; unsigned long _47; signed char _48; _Bool _49; _Bool _50; char * _51; unsigned long _52; unsigned long _53; unsigned long _54; signed char * _55; signed char _56; _Bool _57; unsigned long _58; signed char _59; _Bool _60; _Bool _61; long unsigned int _62; char[10] * _63; unsigned long _64; unsigned long _65; unsigned long _66; signed char * _67; signed char _68; _Bool _69; unsigned long _70; signed char _71; _Bool _72; _Bool _73; char * _74; unsigned long _75; unsigned long _76; unsigned long _77; signed char * _78; signed char _79; _Bool _80; unsigned long _81; signed char _82; _Bool _83; _Bool _84; long unsigned int _85; long unsigned int _86; char * _87; char * _88; unsigned long _89; unsigned long _90; unsigned long _91; signed char * _92; signed char _93; _Bool _94; unsigned long _95; signed char _96; _Bool _97; _Bool _98; char * _99; unsigned long _100; unsigned long _101; unsigned long _102; signed char * _103; signed char _104; _Bool _105; unsigned long _106; signed char _107; _Bool _108; _Bool _109; : foo = {}; _7 = &foo[0]; _8 = (unsigned long) _7; _9 = _8 >> 3; _10 = _9 + 17592186044416; _11 = (signed char *) _10; _12 = *_11; _13 = _12 != 0; _14 = _8 & 7; _15 = (signed char) _14; _16 = _15 >= _12; _17 = _13 & _16; if (_17 != 0) goto ; else goto ; : __asan_report_store1 (_8); : foo[0] = 116; _18 = &foo[1]; _19 = (unsigned long) _18; _20 = _19 >> 3; _21 = _20 + 17592186044416; _22 = (signed char *) _21; _23 = *_22; _24 = _23 != 0; _25 = _19 & 7; _26 = (signed char) _25; _27 = _26 >= _23; _28 = _24 & _27; if (_28 != 0) goto ; else goto ; : __asan_report_store1 (_19); : foo[1] = 101; _29 = &foo[2]; _30 = (unsigned long) _29; _31 = _30 >> 3; _32 = _31 + 17592186044416; _33 = (signed char *) _32; _34 = *_33; _35 = _34 != 0; _36 = _30 & 7; _37 = (signed char) _36; _38 = _37 >= _34; _39 = _35 & _38; if (_39 != 0) goto ; else goto ; : __asan_report_store1 (_30); : foo[2] = 115; _40 = &foo[3]; _41 = (unsigned long) _40; _42 = _41 >> 3; _43 = _42 + 17592186044416; _44 = (signed char *) _43; _45 = *_44; _46 = _45 != 0; _47 = _41 & 7; _48 = (signed char) _47; _49 = _48 >= _45; _50 = _46 & _49; if (_50 != 0) goto ; else goto ; : __asan_report_store1 (_41); : foo[3] = 116; _51 = (char *) &foo; _52 = (unsigned long) _51; _53 = _52 >> 3; _54 = _53 + 17592186044416; _55 = (signed char *) _54; _56 = *_55; _57 = _56 != 0; _58 = _52 & 7; _59 = (signed char) _58; _60 = _59 >= _56; _61 = _57 & _60; if (_61 != 0) goto ; else goto ; : __asan_report_load1 (_52); : _1 = __builtin_strlen (&foo); if (_1 > 0) goto ; else goto ; : _62 = _1 - 1; _63 = _51 + _62; _64 = (unsigned long) _63; _65 = _64 >> 3; _66 = _65 + 17592186044416; _67 = (signed char *) _66; _68 = *_67; _69 = _68 != 0; _70 = _64 & 7; _71 = (signed char) _70; _72 = _71 >= _68; _73 = _69 & _72; if (_73 != 0) goto ; else goto ; : __asan_report_load1 (_64); : : l_2 = (int) _1; n_3 = 10; _4 = n_3 + -4; _5 = (long unsigned int) _4; if (_5 != 0) goto ; else goto ; : _74 = &foo[4]; _75 = (unsigned long) _74; _76 = _75 >> 3; _77 = _76 + 17592186044416; _78 = (signed char *) _77; _79 = *_78; _80 = _79 != 0; _81 = _75 & 7; _82 = (signed char) _81; _83 = _82 >= _79; _84 = _80 & _83; _85 = _5; _86 = _85 - 1; _87 = &foo[4]; _88 = _87 + _86; _89 = (unsigned long) _88; _90 = _89 >> 3; _91 = _90 + 17592186044416; _92 = (signed char *) _91; _93 = *_92; _94 = _93 != 0; _95 = _89 & 7; _96 = (signed char) _95; _97 = _96 >= _93; _98 = _94 & _97; if (_98 != 0) goto ; else goto ; : __asan_report_store1 (_89); : if (_84 != 0) goto ; else goto ; : __asan_report_store1 (_75); : : __builtin_memset (&foo[4], 0, _5); _99 = &foo[11]; _100 = (unsigned long) _99; _101 = _100 >> 3; _102 = _101 + 17592186044416; _103 = (signed char *) _102; _104 = *_103; _105 = _104 != 0; _106 = _100 & 7; _107 = (signed char) _106; _108 = _107 >= _104; _109 = _105 & _108; if (_109 != 0) goto ; else goto ; : __asan_report_store1 (_100); : __sync_fetch_and_add_1 (&foo[11], 1); _6 = l_2; foo ={v} {CLOBBER}; : return _6; } ;; Function _GLOBAL__sub_I_00099_0_foo (_GLOBAL__sub_I_00099_0_foo, funcdef_no=1, decl_uid=1750, cgraph_uid=4) _GLOBAL__sub_I_00099_0_foo () { : __asan_init (); return; } gcc/ * asan.c (insert_if_then_before_iter, instrument_if_then_after_iter, instrument_mem_region_access, instrument_strlen_call, maybe_instrument_builtin_call, maybe_instrument_call): New static functions. (create_cond_insert_point): Renamed create_cond_insert_point_before_iter into this. Add a new parameter to decide whether to insert the condition before or after the statement iterator. (build_check_stmt): Adjust for the new create_cond_insert_point. Add a new parameter to decide whether to add the instrumentation code before or after the statement iterator. (instrument_assignment): Factorize from ... (transform_statements): ... here. Use maybe_instrument_call to instrument builtin function calls as well. (instrument_derefs): Adjust for the new parameter of build_check_stmt. Fix detection of bit-field access. --- gcc/asan.c | 667 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 638 insertions(+), 29 deletions(-) diff --git a/gcc/asan.c b/gcc/asan.c index 39e77e6..89750b0 100644 --- a/gcc/asan.c +++ b/gcc/asan.c @@ -398,9 +398,9 @@ asan_init_func (void) #define PROB_ALWAYS (REG_BR_PROB_BASE) /* Split the current basic block and create a condition statement - insertion point right before the statement pointed to by ITER. - Return an iterator to the point at which the caller might safely - insert the condition statement. + insertion point right before or after the statement pointed to by + ITER. Return an iterator to the point at which the caller might + safely insert the condition statement. THEN_BLOCK must be set to the address of an uninitialized instance of basic_block. The function will then set *THEN_BLOCK to the @@ -414,18 +414,21 @@ asan_init_func (void) statements starting from *ITER, and *THEN_BLOCK is a new empty block. - *ITER is adjusted to still point to the same statement it was - *pointing to initially. */ + *ITER is adjusted to point to always point to the first statement + of the basic block * FALLTHROUGH_BLOCK. That statement is the + same as what ITER was pointing to prior to calling this function, + if BEFORE_P is true; otherwise, it is its following statement. */ static gimple_stmt_iterator -create_cond_insert_point_before_iter (gimple_stmt_iterator *iter, - bool then_more_likely_p, - basic_block *then_block, - basic_block *fallthrough_block) +create_cond_insert_point (gimple_stmt_iterator *iter, + bool before_p, + bool then_more_likely_p, + basic_block *then_block, + basic_block *fallthrough_block) { gimple_stmt_iterator gsi = *iter; - if (!gsi_end_p (gsi)) + if (!gsi_end_p (gsi) && before_p) gsi_prev (&gsi); basic_block cur_bb = gsi_bb (*iter); @@ -466,18 +469,93 @@ create_cond_insert_point_before_iter (gimple_stmt_iterator *iter, return gsi_last_bb (cond_bb); } +/* Insert an if condition followed by a 'then block' right before the + statement pointed to by ITER. The fallthrough block -- which is the + else block of the condition as well as the destination of the + outcoming edge of the 'then block' -- starts with the statement + pointed to by ITER. + + COND is the condition of the if. + + If THEN_MORE_LIKELY_P is true, the probability of the edge to the + 'then block' is higher than the probability of the edge to the + fallthrough block. + + Upon completion of the function, *THEN_BB is set to the newly + inserted 'then block' and similarly, *FALLTHROUGH_BB is set to the + fallthrough block. + + *ITER is adjusted to still point to the same statement it was + pointing to initially. */ + +static void +insert_if_then_before_iter (gimple cond, + gimple_stmt_iterator *iter, + bool then_more_likely_p, + basic_block *then_bb, + basic_block *fallthrough_bb) +{ + gimple_stmt_iterator cond_insert_point = + create_cond_insert_point (iter, + /*before_p=*/true, + then_more_likely_p, + then_bb, + fallthrough_bb); + gsi_insert_after (&cond_insert_point, cond, GSI_NEW_STMT); +} + +/* Insert an if condition followed by a 'then block' right after the + statement pointed to by ITER. The fallthrough block -- which is + the else block of the condition as well as the destination of the + outcoming edge of the 'then block' -- starts with the statement + pointed to by ITER. + + COND is the condition of the if. + + If THEN_MORE_LIKELY_P is true, the probability of the edge to the + 'then block' is higher than the probability of the edge to the + fallthrough block. + + Upon completion of the function, *THEN_BB is set to the newly + inserted 'then block' and similarly, *FALLTHROUGH_BB is set to the + fallthrough block. + + *ITER is adjusted to point to the same statement following the one + it was pointing to prior to calling this function. */ + +static void +insert_if_then_after_iter (gimple cond, + gimple_stmt_iterator *iter, + bool then_more_likely_p, + basic_block *then_bb, + basic_block *fallthrough_bb) +{ + gimple_stmt_iterator cond_insert_point = + create_cond_insert_point (iter, + /*before_p=*/false, + then_more_likely_p, + then_bb, + fallthrough_bb); + gsi_insert_after (&cond_insert_point, cond, GSI_NEW_STMT); +} + /* Instrument the memory access instruction BASE. Insert new - statements before ITER. + statements before or after ITER. Note that the memory access represented by BASE can be either an SSA_NAME, or a non-SSA expression. LOCATION is the source code location. IS_STORE is TRUE for a store, FALSE for a load. - SIZE_IN_BYTES is one of 1, 2, 4, 8, 16. */ + BEFORE_P is TRUE for inserting the instrumentation code before + ITER, FALSE for inserting it after ITER. SIZE_IN_BYTES is one of + 1, 2, 4, 8, 16. + + If BEFORE_P is TRUE, *ITER is arranged to still point to the + statement it was pointing to prior to calling this function, + otherwise, it points to the statement logically following it. */ static void -build_check_stmt (tree base, gimple_stmt_iterator *iter, - location_t location, bool is_store, - int size_in_bytes) +build_check_stmt (location_t location, tree base, gimple_stmt_iterator *iter, + bool before_p, bool is_store, int size_in_bytes) { gimple_stmt_iterator gsi; basic_block then_bb, else_bb; @@ -491,10 +569,10 @@ build_check_stmt (tree base, gimple_stmt_iterator *iter, /* Get an iterator on the point where we can add the condition statement for the instrumentation. */ - gsi = create_cond_insert_point_before_iter (iter, - /*then_more_likely_p=*/false, - &then_bb, - &else_bb); + gsi = create_cond_insert_point (iter, before_p, + /*then_more_likely_p=*/false, + &then_bb, + &else_bb); base = unshare_expr (base); @@ -626,7 +704,7 @@ build_check_stmt (tree base, gimple_stmt_iterator *iter, /* If T represents a memory access, add instrumentation code before ITER. LOCATION is source code location. - IS_STORE is either 1 (for a store) or 0 (for a load). */ + IS_STORE is either TRUE (for a store) or FALSE (for a load). */ static void instrument_derefs (gimple_stmt_iterator *iter, tree t, @@ -661,11 +739,543 @@ instrument_derefs (gimple_stmt_iterator *iter, tree t, int volatilep = 0, unsignedp = 0; get_inner_reference (t, &bitsize, &bitpos, &offset, &mode, &unsignedp, &volatilep, false); - if (bitpos != 0 || bitsize != size_in_bytes * BITS_PER_UNIT) + if (bitpos % (size_in_bytes * BITS_PER_UNIT) + || bitsize != size_in_bytes * BITS_PER_UNIT) return; base = build_fold_addr_expr (t); - build_check_stmt (base, iter, location, is_store, size_in_bytes); + build_check_stmt (location, base, iter, /*before_p=*/true, + is_store, size_in_bytes); +} + +/* Instrument an access to a contiguous memory region that starts at + the address pointed to by BASE, over a length of LEN (expressed in + the sizeof (*BASE) bytes). ITER points to the instruction before + which the instrumentation instructions must be inserted. LOCATION + is the source location that the instrumentation instructions must + have. If IS_STORE is true, then the memory access is a store; + otherwise, it's a load. */ + +static void +instrument_mem_region_access (tree base, tree len, + gimple_stmt_iterator *iter, + location_t location, bool is_store) +{ + if (integer_zerop (len)) + return; + + gimple_stmt_iterator gsi = *iter; + + basic_block fallthrough_bb = NULL, then_bb = NULL; + if (!is_gimple_constant (len)) + { + /* So, the length of the memory area to asan-protect is + non-constant. Let's guard the generated instrumentation code + like: + + if (len != 0) + { + //asan instrumentation code goes here. + } + // falltrough instructions, starting with *ITER. */ + + gimple g = gimple_build_cond (NE_EXPR, + len, + build_int_cst (TREE_TYPE (len), 0), + NULL_TREE, NULL_TREE); + gimple_set_location (g, location); + insert_if_then_before_iter (g, iter, /*then_more_likely_p=*/true, + &then_bb, &fallthrough_bb); + /* Note that fallthrough_bb starts with the statement that was + pointed to by ITER. */ + + /* The 'then block' of the 'if (len != 0) condition is where + we'll generate the asan instrumentation code now. */ + gsi = gsi_start_bb (then_bb); + } + + /* Instrument the beginning of the memory region to be accessed, + and arrange for the rest of the intrumentation code to be + inserted in the then block *after* the current gsi. */ + build_check_stmt (location, base, &gsi, /*before_p=*/true, is_store, 1); + + if (then_bb) + /* We are in the case where the length of the region is not + constant; so instrumentation code is being generated in the + 'then block' of the 'if (len != 0) condition. Let's arrange + for the subsequent instrumentation statements to go in the + 'then block'. */ + gsi = gsi_last_bb (then_bb); + else + *iter = gsi; + + /* We want to instrument the access at the end of the memory region, + which is at (base + len - 1). */ + + /* offset = len - 1; */ + len = unshare_expr (len); + gimple offset = + gimple_build_assign_with_ops (TREE_CODE (len), + make_ssa_name (TREE_TYPE (len), NULL), + len, NULL); + gimple_set_location (offset, location); + gsi_insert_before (&gsi, offset, GSI_NEW_STMT); + + offset = + gimple_build_assign_with_ops (MINUS_EXPR, + make_ssa_name (size_type_node, NULL), + gimple_assign_lhs (offset), + build_int_cst (size_type_node, 1)); + gimple_set_location (offset, location); + gsi_insert_after (&gsi, offset, GSI_NEW_STMT); + + /* _1 = base; */ + base = unshare_expr (base); + gimple region_end = + gimple_build_assign_with_ops (TREE_CODE (base), + make_ssa_name (TREE_TYPE (base), NULL), + base, NULL); + gimple_set_location (region_end, location); + gsi_insert_after (&gsi, region_end, GSI_NEW_STMT); + + /* _2 = _1 + offset; */ + region_end = + gimple_build_assign_with_ops (POINTER_PLUS_EXPR, + make_ssa_name (TREE_TYPE (base), NULL), + gimple_assign_lhs (region_end), + gimple_assign_lhs (offset)); + gimple_set_location (region_end, location); + gsi_insert_after (&gsi, region_end, GSI_NEW_STMT); + + /* instrument access at _2; */ + build_check_stmt (location, gimple_assign_lhs (region_end), + &gsi, /*before_p=*/false, is_store, 1); +} + +/* Instrument the strlen builtin call pointed to by ITER. + + This function instruments the access to the first byte of the + argument, right before the call. After the call it instruments the + access to the last byte of the argument; it uses the result of the + call to deduce the offset of that last byte. */ + +static void +instrument_strlen_call (gimple_stmt_iterator *iter) +{ + gimple call = gsi_stmt (*iter); + gcc_assert (is_gimple_call (call)); + + tree callee = gimple_call_fndecl (call); + gcc_assert (is_builtin_fn (callee) + && DECL_BUILT_IN_CLASS (callee) == BUILT_IN_NORMAL + && DECL_FUNCTION_CODE (callee) == BUILT_IN_STRLEN); + + location_t location = gimple_location (call); + tree str_arg = gimple_call_arg (call, 0); + + /* Instrument the access to the first byte of str_arg. i.e: + + _1 = str_arg; instrument (_1); */ + gimple str_arg_ssa = + gimple_build_assign_with_ops (NOP_EXPR, + make_ssa_name (build_pointer_type + (char_type_node), NULL), + str_arg, NULL); + gimple_set_location (str_arg_ssa, location); + gimple_stmt_iterator gsi = *iter; + gsi_insert_before (&gsi, str_arg_ssa, GSI_NEW_STMT); + build_check_stmt (location, gimple_assign_lhs (str_arg_ssa), &gsi, + /*before_p=*/false, /*is_store=*/false, 1); + + /* If we initially had an instruction like: + + int n = strlen (str) + + we now want to instrument the access to str[n-1], after the + instruction above, if n is greater than 0. */ + + /* So let's build the access to str[n-1], in a if (n > 0) {} block. */ + + tree len = gimple_call_lhs (call); + + /* if (len > 0); */ + gimple cond = gimple_build_cond (GT_EXPR, len, + build_int_cst (size_type_node, 0), + NULL, NULL); + basic_block then_bb, fallthrough_bb; + insert_if_then_after_iter (cond, &gsi, /*then_more_likely_p=*/true, + &then_bb, &fallthrough_bb); + /* fallthrough_bb starts with the statement that gsi pointed to + initially; that is, at the statement that comes after the call + to strlen. The instrumentation code is going to be emitted in + then_bb. */ + + gsi = gsi_start_bb (then_bb); + + /* _3 = _2 - 1; */ + gimple stmt = + gimple_build_assign_with_ops (MINUS_EXPR, + make_ssa_name (size_type_node, NULL), + len, + build_int_cst (size_type_node, 1)); + gimple_set_location (stmt, location); + gsi_insert_after (&gsi, stmt, GSI_NEW_STMT); + + /* really build the access to str[n-1] and instrument it now. That + is, access through the pointer_plus expr: (_1 + _3). */ + stmt = + gimple_build_assign_with_ops (POINTER_PLUS_EXPR, + make_ssa_name (TREE_TYPE (str_arg), + NULL), + gimple_assign_lhs (str_arg_ssa), + gimple_assign_lhs (stmt)); + gimple_set_location (stmt, location); + gsi_insert_after (&gsi, stmt, GSI_NEW_STMT); + + build_check_stmt (location, gimple_assign_lhs (stmt), &gsi, + /*before_p=*/false, /*is_store=*/false, 1); + + /* Ensure that iter points to the statement logically following the + one it was initially pointing to. */ + *iter = gsi_start_bb (fallthrough_bb); +} + +/* if the statement pointed to by the iterator iter is a call to a + builtin memory access function, instrumented it and return true. + otherwise, return false. */ + +static bool +maybe_instrument_builtin_call (gimple_stmt_iterator *iter) +{ + gimple call = gsi_stmt (*iter); + location_t location = gimple_location (call); + + if (!is_gimple_call (call)) + return false; + + tree callee = gimple_call_fndecl (call); + + if (!is_builtin_fn (callee) + || DECL_BUILT_IN_CLASS (callee) != BUILT_IN_NORMAL) + return false; + + tree source0 = NULL_TREE, source1 = NULL_TREE, + dest = NULL_TREE, len = NULL_TREE; + bool is_store = true; + + switch (DECL_FUNCTION_CODE (callee)) + { + /* (s, s, n) style memops. */ + case BUILT_IN_BCMP: + case BUILT_IN_MEMCMP: + len = gimple_call_arg (call, 2); + source0 = gimple_call_arg (call, 0); + source1 = gimple_call_arg (call, 1); + break; + + /* (src, dest, n) style memops. */ + case BUILT_IN_BCOPY: + len = gimple_call_arg (call, 2); + source0 = gimple_call_arg (call, 0); + dest = gimple_call_arg (call, 2); + break; + + /* (dest, src, n) style memops. */ + case BUILT_IN_MEMCPY: + case BUILT_IN_MEMCPY_CHK: + case BUILT_IN_MEMMOVE: + case BUILT_IN_MEMMOVE_CHK: + case BUILT_IN_MEMPCPY: + case BUILT_IN_MEMPCPY_CHK: + dest = gimple_call_arg (call, 0); + source0 = gimple_call_arg (call, 1); + len = gimple_call_arg (call, 2); + break; + + /* (dest, n) style memops. */ + case BUILT_IN_BZERO: + dest = gimple_call_arg (call, 0); + len = gimple_call_arg (call, 1); + break; + + /* (dest, x, n) style memops*/ + case BUILT_IN_MEMSET: + case BUILT_IN_MEMSET_CHK: + dest = gimple_call_arg (call, 0); + len = gimple_call_arg (call, 2); + break; + + case BUILT_IN_STRLEN: + instrument_strlen_call (iter); + return true; + + /* And now the __atomic* and __sync builtins. + These are handled differently from the classical memory memory + access builtins above. */ + + case BUILT_IN_SYNC_FETCH_AND_ADD_1: + case BUILT_IN_SYNC_FETCH_AND_ADD_2: + case BUILT_IN_SYNC_FETCH_AND_ADD_4: + case BUILT_IN_SYNC_FETCH_AND_ADD_8: + case BUILT_IN_SYNC_FETCH_AND_ADD_16: + + case BUILT_IN_SYNC_FETCH_AND_SUB_1: + case BUILT_IN_SYNC_FETCH_AND_SUB_2: + case BUILT_IN_SYNC_FETCH_AND_SUB_4: + case BUILT_IN_SYNC_FETCH_AND_SUB_8: + case BUILT_IN_SYNC_FETCH_AND_SUB_16: + + case BUILT_IN_SYNC_FETCH_AND_OR_1: + case BUILT_IN_SYNC_FETCH_AND_OR_2: + case BUILT_IN_SYNC_FETCH_AND_OR_4: + case BUILT_IN_SYNC_FETCH_AND_OR_8: + case BUILT_IN_SYNC_FETCH_AND_OR_16: + + case BUILT_IN_SYNC_FETCH_AND_AND_1: + case BUILT_IN_SYNC_FETCH_AND_AND_2: + case BUILT_IN_SYNC_FETCH_AND_AND_4: + case BUILT_IN_SYNC_FETCH_AND_AND_8: + case BUILT_IN_SYNC_FETCH_AND_AND_16: + + case BUILT_IN_SYNC_FETCH_AND_XOR_1: + case BUILT_IN_SYNC_FETCH_AND_XOR_2: + case BUILT_IN_SYNC_FETCH_AND_XOR_4: + case BUILT_IN_SYNC_FETCH_AND_XOR_8: + case BUILT_IN_SYNC_FETCH_AND_XOR_16: + + case BUILT_IN_SYNC_FETCH_AND_NAND_1: + case BUILT_IN_SYNC_FETCH_AND_NAND_2: + case BUILT_IN_SYNC_FETCH_AND_NAND_4: + case BUILT_IN_SYNC_FETCH_AND_NAND_8: + + case BUILT_IN_SYNC_ADD_AND_FETCH_1: + case BUILT_IN_SYNC_ADD_AND_FETCH_2: + case BUILT_IN_SYNC_ADD_AND_FETCH_4: + case BUILT_IN_SYNC_ADD_AND_FETCH_8: + case BUILT_IN_SYNC_ADD_AND_FETCH_16: + + case BUILT_IN_SYNC_SUB_AND_FETCH_1: + case BUILT_IN_SYNC_SUB_AND_FETCH_2: + case BUILT_IN_SYNC_SUB_AND_FETCH_4: + case BUILT_IN_SYNC_SUB_AND_FETCH_8: + case BUILT_IN_SYNC_SUB_AND_FETCH_16: + + case BUILT_IN_SYNC_OR_AND_FETCH_1: + case BUILT_IN_SYNC_OR_AND_FETCH_2: + case BUILT_IN_SYNC_OR_AND_FETCH_4: + case BUILT_IN_SYNC_OR_AND_FETCH_8: + case BUILT_IN_SYNC_OR_AND_FETCH_16: + + case BUILT_IN_SYNC_AND_AND_FETCH_1: + case BUILT_IN_SYNC_AND_AND_FETCH_2: + case BUILT_IN_SYNC_AND_AND_FETCH_4: + case BUILT_IN_SYNC_AND_AND_FETCH_8: + case BUILT_IN_SYNC_AND_AND_FETCH_16: + + case BUILT_IN_SYNC_XOR_AND_FETCH_1: + case BUILT_IN_SYNC_XOR_AND_FETCH_2: + case BUILT_IN_SYNC_XOR_AND_FETCH_4: + case BUILT_IN_SYNC_XOR_AND_FETCH_8: + case BUILT_IN_SYNC_XOR_AND_FETCH_16: + + case BUILT_IN_SYNC_NAND_AND_FETCH_1: + case BUILT_IN_SYNC_NAND_AND_FETCH_2: + case BUILT_IN_SYNC_NAND_AND_FETCH_4: + case BUILT_IN_SYNC_NAND_AND_FETCH_8: + + case BUILT_IN_SYNC_BOOL_COMPARE_AND_SWAP_1: + case BUILT_IN_SYNC_BOOL_COMPARE_AND_SWAP_2: + case BUILT_IN_SYNC_BOOL_COMPARE_AND_SWAP_4: + case BUILT_IN_SYNC_BOOL_COMPARE_AND_SWAP_8: + case BUILT_IN_SYNC_BOOL_COMPARE_AND_SWAP_16: + + case BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_1: + case BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_2: + case BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_4: + case BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_8: + case BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_16: + + case BUILT_IN_SYNC_LOCK_TEST_AND_SET_1: + case BUILT_IN_SYNC_LOCK_TEST_AND_SET_2: + case BUILT_IN_SYNC_LOCK_TEST_AND_SET_4: + case BUILT_IN_SYNC_LOCK_TEST_AND_SET_8: + case BUILT_IN_SYNC_LOCK_TEST_AND_SET_16: + + case BUILT_IN_SYNC_LOCK_RELEASE_1: + case BUILT_IN_SYNC_LOCK_RELEASE_2: + case BUILT_IN_SYNC_LOCK_RELEASE_4: + case BUILT_IN_SYNC_LOCK_RELEASE_8: + case BUILT_IN_SYNC_LOCK_RELEASE_16: + + case BUILT_IN_ATOMIC_TEST_AND_SET: + case BUILT_IN_ATOMIC_CLEAR: + case BUILT_IN_ATOMIC_EXCHANGE: + case BUILT_IN_ATOMIC_EXCHANGE_1: + case BUILT_IN_ATOMIC_EXCHANGE_2: + case BUILT_IN_ATOMIC_EXCHANGE_4: + case BUILT_IN_ATOMIC_EXCHANGE_8: + case BUILT_IN_ATOMIC_EXCHANGE_16: + + case BUILT_IN_ATOMIC_LOAD: + case BUILT_IN_ATOMIC_LOAD_1: + case BUILT_IN_ATOMIC_LOAD_2: + case BUILT_IN_ATOMIC_LOAD_4: + case BUILT_IN_ATOMIC_LOAD_8: + case BUILT_IN_ATOMIC_LOAD_16: + + case BUILT_IN_ATOMIC_COMPARE_EXCHANGE: + case BUILT_IN_ATOMIC_COMPARE_EXCHANGE_1: + case BUILT_IN_ATOMIC_COMPARE_EXCHANGE_2: + case BUILT_IN_ATOMIC_COMPARE_EXCHANGE_4: + case BUILT_IN_ATOMIC_COMPARE_EXCHANGE_8: + case BUILT_IN_ATOMIC_COMPARE_EXCHANGE_16: + + case BUILT_IN_ATOMIC_STORE: + case BUILT_IN_ATOMIC_STORE_1: + case BUILT_IN_ATOMIC_STORE_2: + case BUILT_IN_ATOMIC_STORE_4: + case BUILT_IN_ATOMIC_STORE_8: + case BUILT_IN_ATOMIC_STORE_16: + + case BUILT_IN_ATOMIC_ADD_FETCH_1: + case BUILT_IN_ATOMIC_ADD_FETCH_2: + case BUILT_IN_ATOMIC_ADD_FETCH_4: + case BUILT_IN_ATOMIC_ADD_FETCH_8: + case BUILT_IN_ATOMIC_ADD_FETCH_16: + + case BUILT_IN_ATOMIC_SUB_FETCH_1: + case BUILT_IN_ATOMIC_SUB_FETCH_2: + case BUILT_IN_ATOMIC_SUB_FETCH_4: + case BUILT_IN_ATOMIC_SUB_FETCH_8: + case BUILT_IN_ATOMIC_SUB_FETCH_16: + + case BUILT_IN_ATOMIC_AND_FETCH_1: + case BUILT_IN_ATOMIC_AND_FETCH_2: + case BUILT_IN_ATOMIC_AND_FETCH_4: + case BUILT_IN_ATOMIC_AND_FETCH_8: + case BUILT_IN_ATOMIC_AND_FETCH_16: + + case BUILT_IN_ATOMIC_NAND_FETCH_1: + case BUILT_IN_ATOMIC_NAND_FETCH_2: + case BUILT_IN_ATOMIC_NAND_FETCH_4: + case BUILT_IN_ATOMIC_NAND_FETCH_8: + case BUILT_IN_ATOMIC_NAND_FETCH_16: + + case BUILT_IN_ATOMIC_XOR_FETCH_1: + case BUILT_IN_ATOMIC_XOR_FETCH_2: + case BUILT_IN_ATOMIC_XOR_FETCH_4: + case BUILT_IN_ATOMIC_XOR_FETCH_8: + case BUILT_IN_ATOMIC_XOR_FETCH_16: + + case BUILT_IN_ATOMIC_OR_FETCH_1: + case BUILT_IN_ATOMIC_OR_FETCH_2: + case BUILT_IN_ATOMIC_OR_FETCH_4: + case BUILT_IN_ATOMIC_OR_FETCH_8: + case BUILT_IN_ATOMIC_OR_FETCH_16: + + case BUILT_IN_ATOMIC_FETCH_ADD_1: + case BUILT_IN_ATOMIC_FETCH_ADD_2: + case BUILT_IN_ATOMIC_FETCH_ADD_4: + case BUILT_IN_ATOMIC_FETCH_ADD_8: + case BUILT_IN_ATOMIC_FETCH_ADD_16: + + case BUILT_IN_ATOMIC_FETCH_SUB_1: + case BUILT_IN_ATOMIC_FETCH_SUB_2: + case BUILT_IN_ATOMIC_FETCH_SUB_4: + case BUILT_IN_ATOMIC_FETCH_SUB_8: + case BUILT_IN_ATOMIC_FETCH_SUB_16: + + case BUILT_IN_ATOMIC_FETCH_AND_1: + case BUILT_IN_ATOMIC_FETCH_AND_2: + case BUILT_IN_ATOMIC_FETCH_AND_4: + case BUILT_IN_ATOMIC_FETCH_AND_8: + case BUILT_IN_ATOMIC_FETCH_AND_16: + + case BUILT_IN_ATOMIC_FETCH_NAND_1: + case BUILT_IN_ATOMIC_FETCH_NAND_2: + case BUILT_IN_ATOMIC_FETCH_NAND_4: + case BUILT_IN_ATOMIC_FETCH_NAND_8: + case BUILT_IN_ATOMIC_FETCH_NAND_16: + + case BUILT_IN_ATOMIC_FETCH_XOR_1: + case BUILT_IN_ATOMIC_FETCH_XOR_2: + case BUILT_IN_ATOMIC_FETCH_XOR_4: + case BUILT_IN_ATOMIC_FETCH_XOR_8: + case BUILT_IN_ATOMIC_FETCH_XOR_16: + + case BUILT_IN_ATOMIC_FETCH_OR_1: + case BUILT_IN_ATOMIC_FETCH_OR_2: + case BUILT_IN_ATOMIC_FETCH_OR_4: + case BUILT_IN_ATOMIC_FETCH_OR_8: + case BUILT_IN_ATOMIC_FETCH_OR_16: + { + dest = gimple_call_arg (call, 0); + /* So DEST represents the address of a memory location. + instrument_derefs wants the memory location, so lets + dereference the address DEST before handing it to + instrument_derefs. */ + if (TREE_CODE (dest) == ADDR_EXPR) + dest = TREE_OPERAND (dest, 0); + else if (TREE_CODE (dest) == SSA_NAME) + dest = build2 (MEM_REF, TREE_TYPE (TREE_TYPE (dest)), + dest, build_int_cst (TREE_TYPE (dest), 0)); + else + gcc_unreachable (); + + instrument_derefs (iter, dest, location, is_store); + return true; + } + + default: + /* The other builtins memory access are not instrumented in this + function because they either don't have any length parameter, + or their length parameter is just a limit. */ + break; + } + + if (len != NULL_TREE) + { + is_store = (dest != NULL_TREE); + + if (source0 != NULL_TREE) + instrument_mem_region_access (source0, len, iter, + location, is_store); + if (source1 != NULL_TREE) + instrument_mem_region_access (source1, len, iter, + location, is_store); + else if (dest != NULL_TREE) + instrument_mem_region_access (dest, len, iter, + location, is_store); + return true; + } + return false; +} + +/* Instrument the assignment statement ITER if it is subject to + instrumentation. */ + +static void +instrument_assignment (gimple_stmt_iterator *iter) +{ + gimple s = gsi_stmt (*iter); + + gcc_assert (gimple_assign_single_p (s)); + + instrument_derefs (iter, gimple_assign_lhs (s), + gimple_location (s), true); + instrument_derefs (iter, gimple_assign_rhs1 (s), + gimple_location (s), false); +} + +/* Instrument the function call pointed to by the iterator ITER, if it + is subject to instrumentation. At the moment, the only function + calls that are instrumented are some built-in functions that access + memory. Look at maybe_instrument_builtin_call to learn more. */ + +static void +maybe_instrument_call (gimple_stmt_iterator *iter) +{ + maybe_instrument_builtin_call (iter); } /* asan: this looks too complex. Can this be done simpler? */ @@ -686,13 +1296,12 @@ transform_statements (void) if (bb->index >= saved_last_basic_block) continue; for (i = gsi_start_bb (bb); !gsi_end_p (i); gsi_next (&i)) { - gimple s = gsi_stmt (i); - if (!gimple_assign_single_p (s)) - continue; - instrument_derefs (&i, gimple_assign_lhs (s), - gimple_location (s), true); - instrument_derefs (&i, gimple_assign_rhs1 (s), - gimple_location (s), false); + gimple s = gsi_stmt (i); + + if (gimple_assign_single_p (s)) + instrument_assignment (&i); + else if (is_gimple_call (s)) + maybe_instrument_call (&i); } } }