maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/CodeGen/catch-implicit-integer-trun...

343 lines
16 KiB
C
Raw Normal View History

[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// RUN: %clang_cc1 -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s --check-prefix=CHECK
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// RUN: %clang_cc1 -fsanitize=implicit-unsigned-integer-truncation,implicit-signed-integer-truncation -fno-sanitize-recover=implicit-unsigned-integer-truncation,implicit-signed-integer-truncation -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s -implicit-check-not="call void @__ubsan_handle_implicit_conversion" --check-prefixes=CHECK,CHECK-SANITIZE,CHECK-SANITIZE-ANYRECOVER,CHECK-SANITIZE-NORECOVER,CHECK-SANITIZE-UNREACHABLE
// RUN: %clang_cc1 -fsanitize=implicit-unsigned-integer-truncation,implicit-signed-integer-truncation -fsanitize-recover=implicit-unsigned-integer-truncation,implicit-signed-integer-truncation -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s -implicit-check-not="call void @__ubsan_handle_implicit_conversion" --check-prefixes=CHECK,CHECK-SANITIZE,CHECK-SANITIZE-ANYRECOVER,CHECK-SANITIZE-RECOVER
// RUN: %clang_cc1 -fsanitize=implicit-unsigned-integer-truncation,implicit-signed-integer-truncation -fsanitize-trap=implicit-unsigned-integer-truncation,implicit-signed-integer-truncation -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s -implicit-check-not="call void @__ubsan_handle_implicit_conversion" --check-prefixes=CHECK,CHECK-SANITIZE,CHECK-SANITIZE-TRAP,CHECK-SANITIZE-UNREACHABLE
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[UNSIGNED_INT:.*]] = {{.*}} c"'unsigned int'\00" }
// CHECK-SANITIZE-ANYRECOVER: @[[UNSIGNED_CHAR:.*]] = {{.*}} c"'unsigned char'\00" }
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[LINE_100_UNSIGNED_TRUNCATION:.*]] = {{.*}}, i32 100, i32 10 }, {{.*}}* @[[UNSIGNED_INT]], {{.*}}* @[[UNSIGNED_CHAR]], i8 1 }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[SIGNED_INT:.*]] = {{.*}} c"'int'\00" }
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[LINE_200_SIGNED_TRUNCATION:.*]] = {{.*}}, i32 200, i32 10 }, {{.*}}* @[[SIGNED_INT]], {{.*}}* @[[UNSIGNED_CHAR]], i8 2 }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[SIGNED_CHAR:.*]] = {{.*}} c"'signed char'\00" }
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[LINE_300_SIGNED_TRUNCATION:.*]] = {{.*}}, i32 300, i32 10 }, {{.*}}* @[[UNSIGNED_INT]], {{.*}}* @[[SIGNED_CHAR]], i8 2 }
// CHECK-SANITIZE-ANYRECOVER: @[[LINE_400_SIGNED_TRUNCATION:.*]] = {{.*}}, i32 400, i32 10 }, {{.*}}* @[[SIGNED_INT]], {{.*}}* @[[SIGNED_CHAR]], i8 2 }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[UINT32:.*]] = {{.*}} c"'uint32_t' (aka 'unsigned int')\00" }
// CHECK-SANITIZE-ANYRECOVER: @[[UINT8:.*]] = {{.*}} c"'uint8_t' (aka 'unsigned char')\00" }
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-ANYRECOVER: @[[LINE_500_UNSIGNED_TRUNCATION:.*]] = {{.*}}, i32 500, i32 10 }, {{.*}}* @[[UINT32]], {{.*}}* @[[UINT8]], i8 1 }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// ========================================================================== //
// The expected true-positives. These are implicit conversions, and they truncate.
// ========================================================================== //
// CHECK-LABEL: @unsigned_int_to_unsigned_char
unsigned char unsigned_int_to_unsigned_char(unsigned int src) {
// CHECK: %[[DST:.*]] = trunc i32 %[[SRC:.*]] to i8
// CHECK-SANITIZE-NEXT: %[[ANYEXT:.*]] = zext i8 %[[DST]] to i32, !nosanitize
// CHECK-SANITIZE-NEXT: %[[TRUNCHECK:.*]] = icmp eq i32 %[[ANYEXT]], %[[SRC]], !nosanitize
// CHECK-SANITIZE-NEXT: br i1 %[[TRUNCHECK]], label %[[CONT:.*]], label %[[HANDLER_IMPLICIT_CONVERSION:[^,]+]],{{.*}} !nosanitize
// CHECK-SANITIZE: [[HANDLER_IMPLICIT_CONVERSION]]:
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTSRC:.*]] = zext i32 %[[SRC]] to i64, !nosanitize
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTDST:.*]] = zext i8 %[[DST]] to i64, !nosanitize
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-NORECOVER-NEXT: call void @__ubsan_handle_implicit_conversion_abort(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_100_UNSIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
// CHECK-SANITIZE-RECOVER-NEXT: call void @__ubsan_handle_implicit_conversion(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_100_UNSIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-TRAP-NEXT: call void @llvm.trap(){{.*}}, !nosanitize
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-SANITIZE-UNREACHABLE-NEXT: unreachable, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE: [[CONT]]:
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-NEXT: ret i8 %[[DST]]
// CHECK-NEXT: }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
#line 100
return src;
}
// CHECK-LABEL: @signed_int_to_unsigned_char
unsigned char signed_int_to_unsigned_char(signed int src) {
// CHECK: %[[DST:.*]] = trunc i32 %[[SRC:.*]] to i8
// CHECK-SANITIZE-NEXT: %[[ANYEXT:.*]] = zext i8 %[[DST]] to i32, !nosanitize
// CHECK-SANITIZE-NEXT: %[[TRUNCHECK:.*]] = icmp eq i32 %[[ANYEXT]], %[[SRC]], !nosanitize
// CHECK-SANITIZE-NEXT: br i1 %[[TRUNCHECK]], label %[[CONT:.*]], label %[[HANDLER_IMPLICIT_CONVERSION:[^,]+]],{{.*}} !nosanitize
// CHECK-SANITIZE: [[HANDLER_IMPLICIT_CONVERSION]]:
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTSRC:.*]] = zext i32 %[[SRC]] to i64, !nosanitize
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTDST:.*]] = zext i8 %[[DST]] to i64, !nosanitize
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-NORECOVER-NEXT: call void @__ubsan_handle_implicit_conversion_abort(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_200_SIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
// CHECK-SANITIZE-RECOVER-NEXT: call void @__ubsan_handle_implicit_conversion(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_200_SIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-TRAP-NEXT: call void @llvm.trap(){{.*}}, !nosanitize
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-SANITIZE-UNREACHABLE-NEXT: unreachable, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE: [[CONT]]:
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-NEXT: ret i8 %[[DST]]
// CHECK-NEXT: }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
#line 200
return src;
}
// CHECK-LABEL: @unsigned_int_to_signed_char
signed char unsigned_int_to_signed_char(unsigned int src) {
// CHECK: %[[DST:.*]] = trunc i32 %[[SRC:.*]] to i8
// CHECK-SANITIZE-NEXT: %[[ANYEXT:.*]] = sext i8 %[[DST]] to i32, !nosanitize
// CHECK-SANITIZE-NEXT: %[[TRUNCHECK:.*]] = icmp eq i32 %[[ANYEXT]], %[[SRC]], !nosanitize
// CHECK-SANITIZE-NEXT: br i1 %[[TRUNCHECK]], label %[[CONT:.*]], label %[[HANDLER_IMPLICIT_CONVERSION:[^,]+]],{{.*}} !nosanitize
// CHECK-SANITIZE: [[HANDLER_IMPLICIT_CONVERSION]]:
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTSRC:.*]] = zext i32 %[[SRC]] to i64, !nosanitize
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTDST:.*]] = zext i8 %[[DST]] to i64, !nosanitize
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-NORECOVER-NEXT: call void @__ubsan_handle_implicit_conversion_abort(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_300_SIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
// CHECK-SANITIZE-RECOVER-NEXT: call void @__ubsan_handle_implicit_conversion(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_300_SIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-TRAP-NEXT: call void @llvm.trap(){{.*}}, !nosanitize
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-SANITIZE-UNREACHABLE-NEXT: unreachable, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE: [[CONT]]:
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-NEXT: ret i8 %[[DST]]
// CHECK-NEXT: }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
#line 300
return src;
}
// CHECK-LABEL: @signed_int_to_signed_char
signed char signed_int_to_signed_char(signed int src) {
// CHECK: %[[DST:.*]] = trunc i32 %[[SRC:.*]] to i8
// CHECK-SANITIZE-NEXT: %[[ANYEXT:.*]] = sext i8 %[[DST]] to i32, !nosanitize
// CHECK-SANITIZE-NEXT: %[[TRUNCHECK:.*]] = icmp eq i32 %[[ANYEXT]], %[[SRC]], !nosanitize
// CHECK-SANITIZE-NEXT: br i1 %[[TRUNCHECK]], label %[[CONT:.*]], label %[[HANDLER_IMPLICIT_CONVERSION:[^,]+]],{{.*}} !nosanitize
// CHECK-SANITIZE: [[HANDLER_IMPLICIT_CONVERSION]]:
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTSRC:.*]] = zext i32 %[[SRC]] to i64, !nosanitize
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTDST:.*]] = zext i8 %[[DST]] to i64, !nosanitize
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-NORECOVER-NEXT: call void @__ubsan_handle_implicit_conversion_abort(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_400_SIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
// CHECK-SANITIZE-RECOVER-NEXT: call void @__ubsan_handle_implicit_conversion(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_400_SIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-TRAP-NEXT: call void @llvm.trap(){{.*}}, !nosanitize
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-SANITIZE-UNREACHABLE-NEXT: unreachable, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE: [[CONT]]:
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-NEXT: ret i8 %[[DST]]
// CHECK-NEXT: }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
#line 400
return src;
}
// ========================================================================== //
// Check canonical type stuff
// ========================================================================== //
typedef unsigned int uint32_t;
typedef unsigned char uint8_t;
// CHECK-LABEL: @uint32_to_uint8
uint8_t uint32_to_uint8(uint32_t src) {
// CHECK: %[[DST:.*]] = trunc i32 %[[SRC:.*]] to i8
// CHECK-SANITIZE-NEXT: %[[ANYEXT:.*]] = zext i8 %[[DST]] to i32, !nosanitize
// CHECK-SANITIZE-NEXT: %[[TRUNCHECK:.*]] = icmp eq i32 %[[ANYEXT]], %[[SRC]], !nosanitize
// CHECK-SANITIZE-NEXT: br i1 %[[TRUNCHECK]], label %[[CONT:.*]], label %[[HANDLER_IMPLICIT_CONVERSION:[^,]+]],{{.*}} !nosanitize
// CHECK-SANITIZE: [[HANDLER_IMPLICIT_CONVERSION]]:
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTSRC:.*]] = zext i32 %[[SRC]] to i64, !nosanitize
// CHECK-SANITIZE-ANYRECOVER-NEXT: %[[EXTDST:.*]] = zext i8 %[[DST]] to i64, !nosanitize
[clang][ubsan] Split Implicit Integer Truncation Sanitizer into unsigned and signed checks Summary: As per IRC disscussion, it seems we really want to have more fine-grained `-fsanitize=implicit-integer-truncation`: * A check when both of the types are unsigned. * Another check for the other cases (either one of the types is signed, or both of the types is signed). This is clang part. Compiler-rt part is D50902. Reviewers: rsmith, vsk, Sanitizers Reviewed by: rsmith Differential Revision: https://reviews.llvm.org/D50901 llvm-svn: 344230
2018-10-11 17:09:50 +08:00
// CHECK-SANITIZE-NORECOVER-NEXT: call void @__ubsan_handle_implicit_conversion_abort(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_500_UNSIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
// CHECK-SANITIZE-RECOVER-NEXT: call void @__ubsan_handle_implicit_conversion(i8* bitcast ({ {{{.*}}}, {{{.*}}}*, {{{.*}}}*, i8 }* @[[LINE_500_UNSIGNED_TRUNCATION]] to i8*), i64 %[[EXTSRC]], i64 %[[EXTDST]]){{.*}}, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE-TRAP-NEXT: call void @llvm.trap(){{.*}}, !nosanitize
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-SANITIZE-UNREACHABLE-NEXT: unreachable, !nosanitize
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
// CHECK-SANITIZE: [[CONT]]:
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
// CHECK-NEXT: ret i8 %[[DST]]
// CHECK-NEXT: }
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
#line 500
return src;
}
// ========================================================================== //
// Check that explicit conversion does not interfere with implicit conversion
// ========================================================================== //
// These contain one implicit truncating conversion, and one explicit truncating conversion.
// We want to make sure that we still diagnose the implicit conversion.
// Implicit truncation after explicit truncation.
// CHECK-LABEL: @explicit_conversion_interference0
unsigned char explicit_conversion_interference0(unsigned int c) {
// CHECK-SANITIZE: %[[ANYEXT:.*]] = zext i8 %[[DST:.*]] to i16, !nosanitize
// CHECK-SANITIZE: call
return (unsigned short)c;
}
// Implicit truncation before explicit truncation.
// CHECK-LABEL: @explicit_conversion_interference1
unsigned char explicit_conversion_interference1(unsigned int c) {
// CHECK-SANITIZE: %[[ANYEXT:.*]] = zext i16 %[[DST:.*]] to i32, !nosanitize
// CHECK-SANITIZE: call
unsigned short b;
return (unsigned char)(b = c);
}
// ========================================================================== //
// The expected true-negatives.
// ========================================================================== //
// Sanitization is explicitly disabled.
// ========================================================================== //
// CHECK-LABEL: @blacklist_0
__attribute__((no_sanitize("undefined"))) unsigned char blacklist_0(unsigned int src) {
// We are not in "undefined" group, so that doesn't work.
// CHECK-SANITIZE: call
return src;
}
// CHECK-LABEL: @blacklist_1
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
__attribute__((no_sanitize("integer"))) unsigned char blacklist_1(unsigned int src) {
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
return src;
}
// CHECK-LABEL: @blacklist_2
[NFC] Some small test updates for Implicit Conversion sanitizer. Split off from D50250. llvm-svn: 339995
2018-08-17 15:33:25 +08:00
__attribute__((no_sanitize("implicit-conversion"))) unsigned char blacklist_2(unsigned int src) {
return src;
}
// CHECK-LABEL: @blacklist_3
__attribute__((no_sanitize("implicit-integer-truncation"))) unsigned char blacklist_3(unsigned int src) {
[clang][ubsan] Implicit Conversion Sanitizer - integer truncation - clang part Summary: C and C++ are interesting languages. They are statically typed, but weakly. The implicit conversions are allowed. This is nice, allows to write code while balancing between getting drowned in everything being convertible, and nothing being convertible. As usual, this comes with a price: ``` unsigned char store = 0; bool consume(unsigned int val); void test(unsigned long val) { if (consume(val)) { // the 'val' is `unsigned long`, but `consume()` takes `unsigned int`. // If their bit widths are different on this platform, the implicit // truncation happens. And if that `unsigned long` had a value bigger // than UINT_MAX, then you may or may not have a bug. // Similarly, integer addition happens on `int`s, so `store` will // be promoted to an `int`, the sum calculated (0+768=768), // and the result demoted to `unsigned char`, and stored to `store`. // In this case, the `store` will still be 0. Again, not always intended. store = store + 768; // before addition, 'store' was promoted to int. } // But yes, sometimes this is intentional. // You can either make the conversion explicit (void)consume((unsigned int)val); // or mask the value so no bits will be *implicitly* lost. (void)consume((~((unsigned int)0)) & val); } ``` Yes, there is a `-Wconversion`` diagnostic group, but first, it is kinda noisy, since it warns on everything (unlike sanitizers, warning on an actual issues), and second, there are cases where it does **not** warn. So a Sanitizer is needed. I don't have any motivational numbers, but i know i had this kind of problem 10-20 times, and it was never easy to track down. The logic to detect whether an truncation has happened is pretty simple if you think about it - https://godbolt.org/g/NEzXbb - basically, just extend (using the new, not original!, signedness) the 'truncated' value back to it's original width, and equality-compare it with the original value. The most non-trivial thing here is the logic to detect whether this `ImplicitCastExpr` AST node is **actually** an implicit conversion, //or// part of an explicit cast. Because the explicit casts are modeled as an outer `ExplicitCastExpr` with some `ImplicitCastExpr`'s as **direct** children. https://godbolt.org/g/eE1GkJ Nowadays, we can just use the new `part_of_explicit_cast` flag, which is set on all the implicitly-added `ImplicitCastExpr`'s of an `ExplicitCastExpr`. So if that flag is **not** set, then it is an actual implicit conversion. As you may have noted, this isn't just named `-fsanitize=implicit-integer-truncation`. There are potentially some more implicit conversions to be warned about. Namely, implicit conversions that result in sign change; implicit conversion between different floating point types, or between fp and an integer, when again, that conversion is lossy. One thing i know isn't handled is bitfields. This is a clang part. The compiler-rt part is D48959. Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=21530 | PR21530 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=37552 | PR37552 ]], [[ https://bugs.llvm.org/show_bug.cgi?id=35409 | PR35409 ]]. Partially fixes [[ https://bugs.llvm.org/show_bug.cgi?id=9821 | PR9821 ]]. Fixes https://github.com/google/sanitizers/issues/940. (other than sign-changing implicit conversions) Reviewers: rjmccall, rsmith, samsonov, pcc, vsk, eugenis, efriedma, kcc, erichkeane Reviewed By: rsmith, vsk, erichkeane Subscribers: erichkeane, klimek, #sanitizers, aaron.ballman, RKSimon, dtzWill, filcab, danielaustin, ygribov, dvyukov, milianw, mclow.lists, cfe-commits, regehr Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D48958 llvm-svn: 338288
2018-07-31 02:58:30 +08:00
return src;
}
// Explicit truncating conversions.
// ========================================================================== //
// CHECK-LABEL: @explicit_unsigned_int_to_unsigned_char
unsigned char explicit_unsigned_int_to_unsigned_char(unsigned int src) {
return (unsigned char)src;
}
// CHECK-LABEL: @explicit_signed_int_to_unsigned_char
unsigned char explicit_signed_int_to_unsigned_char(signed int src) {
return (unsigned char)src;
}
// CHECK-LABEL: @explicit_unsigned_int_to_signed_char
signed char explicit_unsigned_int_to_signed_char(unsigned int src) {
return (signed char)src;
}
// CHECK-LABEL: @explicit_signed_int_to_signed_char
signed char explicit_signed_int_to_signed_char(signed int src) {
return (signed char)src;
}
// Explicit NOP conversions.
// ========================================================================== //
// CHECK-LABEL: @explicit_unsigned_int_to_unsigned_int
unsigned int explicit_unsigned_int_to_unsigned_int(unsigned int src) {
return (unsigned int)src;
}
// CHECK-LABEL: @explicit_signed_int_to_signed_int
signed int explicit_signed_int_to_signed_int(signed int src) {
return (signed int)src;
}
// CHECK-LABEL: @explicit_unsigned_char_to_signed_char
unsigned char explicit_unsigned_char_to_signed_char(unsigned char src) {
return (unsigned char)src;
}
// CHECK-LABEL: @explicit_signed_char_to_signed_char
signed char explicit_signed_char_to_signed_char(signed char src) {
return (signed char)src;
}
// upcasts.
// ========================================================================== //
// CHECK-LABEL: @unsigned_char_to_unsigned_int
unsigned int unsigned_char_to_unsigned_int(unsigned char src) {
return src;
}
// CHECK-LABEL: @signed_char_to_unsigned_int
unsigned int signed_char_to_unsigned_int(signed char src) {
return src;
}
// CHECK-LABEL: @unsigned_char_to_signed_int
signed int unsigned_char_to_signed_int(unsigned char src) {
return src;
}
// CHECK-LABEL: @signed_char_to_signed_int
signed int signed_char_to_signed_int(signed char src) {
return src;
}
// Explicit upcasts.
// ========================================================================== //
// CHECK-LABEL: @explicit_unsigned_char_to_unsigned_int
unsigned int explicit_unsigned_char_to_unsigned_int(unsigned char src) {
return (unsigned int)src;
}
// CHECK-LABEL: @explicit_signed_char_to_unsigned_int
unsigned int explicit_signed_char_to_unsigned_int(signed char src) {
return (unsigned int)src;
}
// CHECK-LABEL: @explicit_unsigned_char_to_signed_int
signed int explicit_unsigned_char_to_signed_int(unsigned char src) {
return (signed int)src;
}
// CHECK-LABEL: @explicit_signed_char_to_signed_int
signed int explicit_signed_char_to_signed_int(signed char src) {
return (signed int)src;
}
// conversions to to boolean type are not counted as truncation.
// ========================================================================== //
// CHECK-LABEL: @unsigned_int_to_bool
_Bool unsigned_int_to_bool(unsigned int src) {
return src;
}
// CHECK-LABEL: @signed_int_to_bool
_Bool signed_int_to_bool(signed int src) {
return src;
}
// CHECK-LABEL: @explicit_unsigned_int_to_bool
_Bool explicit_unsigned_int_to_bool(unsigned int src) {
return (_Bool)src;
}
// CHECK-LABEL: @explicit_signed_int_to_bool
_Bool explicit_signed_int_to_bool(signed int src) {
return (_Bool)src;
}
// Explicit truncating conversions from pointer to a much-smaller integer.
// Can not have an implicit conversion from pointer to an integer.
// Can not have an implicit conversion between two enums.
// ========================================================================== //
// CHECK-LABEL: @explicit_voidptr_to_unsigned_char
unsigned char explicit_voidptr_to_unsigned_char(void *src) {
return (unsigned char)src;
}
// CHECK-LABEL: @explicit_voidptr_to_signed_char
signed char explicit_voidptr_to_signed_char(void *src) {
return (signed char)src;
}
// Implicit truncating conversions from floating-point may result in precision loss.
// ========================================================================== //
// CHECK-LABEL: @float_to_unsigned_int
unsigned int float_to_unsigned_int(float src) {
return src;
}
// CHECK-LABEL: @float_to_signed_int
signed int float_to_signed_int(float src) {
return src;
}
// CHECK-LABEL: @double_to_unsigned_int
unsigned int double_to_unsigned_int(double src) {
return src;
}
// CHECK-LABEL: @double_to_signed_int
signed int double_to_signed_int(double src) {
return src;
}
// Implicit truncating conversions between fp may result in precision loss.
// ========================================================================== //
// CHECK-LABEL: @double_to_float
float double_to_float(double src) {
return src;
}