include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++----- tools/include/linux/unaligned.h | 4 --- tools/include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++----- 3 files changed, 84 insertions(+), 16 deletions(-)
The existing type punning approach with packed structs requires
-fno-strict-aliasing to be passed to the compiler for
correctness. This is true in the kernel tree but not in the tools
directory resulting in this suggested patch from Eric Biggers
<ebiggers@google.com>:
https://lore.kernel.org/lkml/20250614044133.660848-2-ebiggers@kernel.org/
Requiring -fno-strict-aliasing seems unfortunate and so this patch
makes the unaligned code work via memcpy for type punning rather than
the packed attribute.
v2: switch memcpy to __builtin_memcpy to avoid potential/disallowed
memcpy calls in vdso caused by -fno-builtin. Reported by
Christophe Leroy <christophe.leroy@csgroup.eu>:
https://lore.kernel.org/lkml/c57de5bf-d55c-48c5-9dfa-e2fb844dafe9@csgroup.eu/
Ian Rogers (3):
vdso: Switch get/put unaligned from packed struct to memcpy
tools headers: Update the linux/unaligned.h copy with the kernel
sources
tools headers: Remove unneeded ignoring of warnings in unaligned.h
include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++-----
tools/include/linux/unaligned.h | 4 ---
tools/include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++-----
3 files changed, 84 insertions(+), 16 deletions(-)
--
2.50.0.rc2.701.gf1e915cc24-goog
Le 17/06/2025 à 22:53, Ian Rogers a écrit : > The existing type punning approach with packed structs requires > -fno-strict-aliasing to be passed to the compiler for > correctness. This is true in the kernel tree but not in the tools > directory resulting in this suggested patch from Eric Biggers > <ebiggers@google.com>: > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Flkml%2F20250614044133.660848-2-ebiggers%40kernel.org%2F&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7Cf05413010ecc40ad1bdf08ddade1316a%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C638857904894967529%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=29QUBLnUowncZiTH4z74Ec1olUlX0OTYnUNGDvWxX1o%3D&reserved=0 > > Requiring -fno-strict-aliasing seems unfortunate and so this patch > makes the unaligned code work via memcpy for type punning rather than > the packed attribute. > > v2: switch memcpy to __builtin_memcpy to avoid potential/disallowed > memcpy calls in vdso caused by -fno-builtin. Reported by > Christophe Leroy <christophe.leroy@csgroup.eu>: > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Flkml%2Fc57de5bf-d55c-48c5-9dfa-e2fb844dafe9%40csgroup.eu%2F&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7Cf05413010ecc40ad1bdf08ddade1316a%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C638857904894985987%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=l1LJP3XPKVuhDpiHxxkfWBiNPMYaBT9YXGBFzs6wLpY%3D&reserved=0 > Does this new version also fixes the failures reported by the robots: - arm64-randconfig with clang 21 - i386-randconfig with gcc 12 Christophe > Ian Rogers (3): > vdso: Switch get/put unaligned from packed struct to memcpy > tools headers: Update the linux/unaligned.h copy with the kernel > sources > tools headers: Remove unneeded ignoring of warnings in unaligned.h > > include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++----- > tools/include/linux/unaligned.h | 4 --- > tools/include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++----- > 3 files changed, 84 insertions(+), 16 deletions(-) >
On Wed, Jun 18, 2025 at 4:42 AM Christophe Leroy <christophe.leroy@csgroup.eu> wrote: > > > > Le 17/06/2025 à 22:53, Ian Rogers a écrit : > > The existing type punning approach with packed structs requires > > -fno-strict-aliasing to be passed to the compiler for > > correctness. This is true in the kernel tree but not in the tools > > directory resulting in this suggested patch from Eric Biggers > > <ebiggers@google.com>: > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Flkml%2F20250614044133.660848-2-ebiggers%40kernel.org%2F&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7Cf05413010ecc40ad1bdf08ddade1316a%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C638857904894967529%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=29QUBLnUowncZiTH4z74Ec1olUlX0OTYnUNGDvWxX1o%3D&reserved=0 > > > > Requiring -fno-strict-aliasing seems unfortunate and so this patch > > makes the unaligned code work via memcpy for type punning rather than > > the packed attribute. > > > > v2: switch memcpy to __builtin_memcpy to avoid potential/disallowed > > memcpy calls in vdso caused by -fno-builtin. Reported by > > Christophe Leroy <christophe.leroy@csgroup.eu>: > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Flkml%2Fc57de5bf-d55c-48c5-9dfa-e2fb844dafe9%40csgroup.eu%2F&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7Cf05413010ecc40ad1bdf08ddade1316a%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C638857904894985987%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=l1LJP3XPKVuhDpiHxxkfWBiNPMYaBT9YXGBFzs6wLpY%3D&reserved=0 > > So I think I can remove the _Generic block in unaligned.h by depending on the equivalent __unqual_scalar_typeof from include/linux/compiler_types.h. I think this is preferable and so I'll post a v3. > Does this new version also fixes the failures reported by the robots: > - arm64-randconfig with clang 21 The issue here was: llvm-readelf: warning: 'arch/arm64/kernel/vdso/vdso.so.dbg': invalid PT_DYNAMIC size (0x118) llvm-readelf: warning: 'arch/arm64/kernel/vdso/vdso.so.dbg': PT_DYNAMIC dynamic table is invalid: SHT_DYNAMIC will be used which the switch to __builtin_memcpy should have resolved. > - i386-randconfig with gcc 12 The reported error is in the unaligned code from sparse: drivers/scsi/megaraid/megaraid_sas_base.c:8438:32: sparse: sparse: Using plain integer as NULL pointer I believe this is because of the _Generic block with a pointer type, and so the use of __unqual_scalar_typeof in the next/v3 patch set should hopefully resolve it - or else we'll need to special case pointer types to assign NULL rather than 0, but I'd prefer not to special case every potential pointer type. Thanks, Ian > Christophe > > > Ian Rogers (3): > > vdso: Switch get/put unaligned from packed struct to memcpy > > tools headers: Update the linux/unaligned.h copy with the kernel > > sources > > tools headers: Remove unneeded ignoring of warnings in unaligned.h > > > > include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++----- > > tools/include/linux/unaligned.h | 4 --- > > tools/include/vdso/unaligned.h | 48 ++++++++++++++++++++++++++++----- > > 3 files changed, 84 insertions(+), 16 deletions(-) > > >
© 2016 - 2025 Red Hat, Inc.