__aeabi_uldivmod in float64_rem() after LLVM commit 3bc09c7da50a7bb2df51b1829f837f017da5ee34 #1666
Assignees
Labels
[BUG] linux
A bug that should be fixed in the mainline kernel.
[BUG] llvm (main)
A bug in an unreleased version of LLVM (this label is appropriate for regressions)
Reported upstream
This bug was filed on LLVM’s issue tracker, Phabricator, or the kernel mailing list.
[WORKAROUND] Applied
This bug has an applied workaround
Comments
nathanchance
added
[BUG] Untriaged
|
cc @preames |
nickdesaulniers
added
the
Reported upstream
Closed
nathanchance
added a commit
to nathanchance/continuous-integration2
that referenced
this issue
Aug 1, 2022
There is a compiler inserted call to __aeabi_uldivmod, which causes an error at link time because the kernel does not link against compiler-rt. Disable this configuration until it can be dealt with. Link: ClangBuiltLinux/linux#1666 Signed-off-by: Nathan Chancellor <nathan@kernel.org>
|
AFAICT, there is no division by constant here; so https://reviews.llvm.org/D130862 won't/doesn't help. |
|
Android is hitting this now with https://android-review.googlesource.com/c/kernel/common/+/2237552. |
|
So it looks like https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm/nwfpe/softfloat.c is based on this library: http://www.jhauser.us/arithmetic/SoftFloat.html. Specifically, it seems like what's in tree is an old version of the version 2 release. Looks like the latest release is version 3 which looks like a rewrite, but for version 2 there's a version 2c that has an updated implementation of EDIT: Oh, maybe /tmp/SoftFloat-2c/softfloat/bits64/softfloat.c is the version the kernel is using? Now called |
diff --git a/arch/arm/nwfpe/softfloat.c b/arch/arm/nwfpe/softfloat.c
index ffa6b438786b..a0a7bae0b8b9 100644
--- a/arch/arm/nwfpe/softfloat.c
+++ b/arch/arm/nwfpe/softfloat.c
@@ -2309,7 +2309,7 @@ float64 float64_rem( struct roundingData *roundData, float64 a, float64 b )
do {
alternateASig = aSig;
++q;
- aSig -= bSig;
+ /*aSig -= bSig;*/
} while ( 0 <= (sbits64) aSig );
sigMean = aSig + alternateASig;
if ( ( sigMean < 0 ) || ( ( sigMean == 0 ) && ( q & 1 ) ) ) {makes the |
|
submitted to RMK's queue: https://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=9256/1 |
nickdesaulniers
added
[BUG] linux
nickdesaulniers
added a commit
to ClangBuiltLinux/continuous-integration2
that referenced
this issue
Oct 13, 2022
These are not necessary in android14-5.15 or android-mainline because I have already merged them there. #404 ClangBuiltLinux/linux#1666 https://android-review.googlesource.com/c/kernel/common/+/2250393/2 https://android-review.googlesource.com/c/kernel/common/+/2249815/2
gatieme
pushed a commit
to gatieme/aosp_kernel_common
that referenced
this issue
Oct 24, 2022
clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Bug: 250930332 Link: https://lore.kernel.org/lkml/20221010225342.3903590-1-ndesaulniers@google.com/ Change-Id: I038f06e43420e72927956133a8fe39a3db21a9db
gatieme
pushed a commit
to gatieme/aosp_kernel_common
that referenced
this issue
Oct 24, 2022
clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Bug: 250930332 Link: https://lore.kernel.org/lkml/20221010225342.3903590-1-ndesaulniers@google.com/ Change-Id: I038f06e43420e72927956133a8fe39a3db21a9db
nickdesaulniers
added
[PATCH] Accepted
jonhunter
pushed a commit
to jonhunter/linux
that referenced
this issue
Nov 8, 2022
clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
|
The workaround appears to be accepted: https://git.kernel.org/next/linux-next/c/3220022038b9a3845eea762af85f1c5694b9f861 |
|
The workaround has been merged into mainline: https://git.kernel.org/linus/3220022038b9a3845eea762af85f1c5694b9f861 |
jgudec
pushed a commit
to jgudec/android_kernel_samsung_exynos2200
that referenced
this issue
Jul 4, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
codewalkerster
pushed a commit
to hardkernel/linux
that referenced
this issue
Jul 5, 2023
commit 3220022 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ahnet-69
pushed a commit
to ahnet-69/android_kernel_samsung_a32
that referenced
this issue
Jul 17, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Santhanabalan
pushed a commit
to Santhanabalan/kernel_xiaomi_sm8350
that referenced
this issue
Jul 21, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
REV3NT3CH
pushed a commit
to B14CKB1RD-Kernel/B14CKB1RD_floral
that referenced
this issue
Jul 25, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
LucasBlackLu
pushed a commit
to LucasBlackLu/kernel_google_msm-4.14
that referenced
this issue
Jul 25, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mizdrake7
pushed a commit
to mizdrake7/Graveyard_r5x
that referenced
this issue
Aug 2, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mizdrake7
pushed a commit
to mizdrake7/Graveyard_r5x
that referenced
this issue
Aug 2, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mizdrake7
pushed a commit
to mizdrake7/Graveyard_r5x
that referenced
this issue
Aug 4, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
iMxNizam
pushed a commit
to iMxNizam/kernel_xiaomi_laurel_sprout
that referenced
this issue
Aug 6, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
wanghao75
pushed a commit
to openeuler-mirror/kernel
that referenced
this issue
Aug 14, 2023
stable inclusion from stable-v5.10.163 commit 1032520b2109d17980ed21135aac891d995f7206 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I7PJ9N Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=1032520b2109d17980ed21135aac891d995f7206 ---------------------------------------------------- commit 3220022 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: zhaoxiaoqiang11 <zhaoxiaoqiang11@jd.com>
wanghao75
pushed a commit
to openeuler-mirror/kernel
that referenced
this issue
Aug 26, 2023
stable inclusion from stable-v5.10.163 commit 1032520b2109d17980ed21135aac891d995f7206 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I7PJ9N Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=1032520b2109d17980ed21135aac891d995f7206 ---------------------------------------------------- commit 3220022 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: zhaoxiaoqiang11 <zhaoxiaoqiang11@jd.com>
log1cs
pushed a commit
to log1cs/kernel_nokia_msm8998
that referenced
this issue
Sep 19, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ulrich Hecht <uli+cip@fpond.eu>
ratatouille100
pushed a commit
to ratatouille100/kernel_samsung_universal9611
that referenced
this issue
Dec 19, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
lineageos-gerrit
pushed a commit
to LineageOS/android_kernel_motorola_sdm632
that referenced
this issue
Dec 31, 2023
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
vullk4n
pushed a commit
to vullk4n/kernel_motorola_sdm632
that referenced
this issue
Jan 8, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
backslashxx
pushed a commit
to backslashxx/android_karnol_ximi_fog
that referenced
this issue
Feb 15, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
backslashxx
pushed a commit
to backslashxx/android_karnol_ximi_fog
that referenced
this issue
Feb 15, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
LinuxGuy312
pushed a commit
to LinuxGuy312/android_kernel_realme_RMX1805
that referenced
this issue
Apr 8, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
aled99
pushed a commit
to aled99/kernel_xiaomi_sm8250
that referenced
this issue
Apr 8, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Huawei-Dev
pushed a commit
to Huawei-Dev/android_kernel_huawei_hi3660
that referenced
this issue
Apr 13, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ALprjkt
pushed a commit
to ALprjkt/kernel_xiaomi_ysl
that referenced
this issue
May 18, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ALprjkt
pushed a commit
to ALprjkt/kernel_xiaomi_ysl
that referenced
this issue
May 20, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Huawei-Dev
pushed a commit
to Huawei-Dev/android_kernel_huawei_hi3660
that referenced
this issue
May 23, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
user-why-red
pushed a commit
to user-why-red/android_kernel_xiaomi_sdm660_44
that referenced
this issue
May 24, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Santhosh <santhosh.user.why.red@gmail.com>
sidex15
pushed a commit
to sidex15/android_kernel_lge_sm8150
that referenced
this issue
May 25, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
user-why-red
pushed a commit
to user-why-red/android_kernel_xiaomi_sdm660_44
that referenced
this issue
May 27, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Santhosh <santhosh.user.why.red@gmail.com>
kevios12
pushed a commit
to kevios12/android_kernel_samsung_universal7885
that referenced
this issue
Jun 10, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ulrich Hecht <uli+cip@fpond.eu>
aled99
pushed a commit
to aled99/kernel_xiaomi_sm8250
that referenced
this issue
Jun 10, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
garry-rogov
pushed a commit
to garry-rogov/android_kernel_xiaomi_garnet
that referenced
this issue
Jul 21, 2024
commit 3220022038b9a3845eea762af85f1c5694b9f861 upstream. clang-15's ability to elide loops completely became more aggressive when it can deduce how a variable is being updated in a loop. Counting down one variable by an increment of another can be replaced by a modulo operation. For 64b variables on 32b ARM EABI targets, this can result in the compiler generating calls to __aeabi_uldivmod, which it does for a do while loop in float64_rem(). For the kernel, we'd generally prefer that developers not open code 64b division via binary / operators and instead use the more explicit helpers from div64.h. On arm-linux-gnuabi targets, failure to do so can result in linkage failures due to undefined references to __aeabi_uldivmod(). While developers can avoid open coding divisions on 64b variables, the compiler doesn't know that the Linux kernel has a partial implementation of a compiler runtime (--rtlib) to enforce this convention. It's also undecidable for the compiler whether the code in question would be faster to execute the loop vs elide it and do the 64b division. While I actively avoid using the internal -mllvm command line flags, I think we get better code than using barrier() here, which will force reloads+spills in the loop for all toolchains. Link: ClangBuiltLinux/linux#1666 Reported-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
After llvm/llvm-project@3bc09c7, I see the following error while building
ARCH=arm allmodconfig:arch/arm/nwfpe/softfloat.cis kind of dense, I reduced out the following usingcvise:The text was updated successfully, but these errors were encountered: