Skip to contentSkip to conversation

kdev_t: Always inline major/minor helper functions

PATCH Applied Josh Poimboeuf 3 participants 9 messages Dec 23, 2020 – Jan 4, 2021 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 kdev_t: Always inline major/minor helper functions

Silly GCC doesn't always inline these trivial functions.

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

---
 include/linux/kdev_t.h | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)
include/linux/kdev_t.h +11 -11
index 85b5151911cf..4856706fbfeb 100644
@@ -21,61 +21,61 @@
 	})
 
 /* acceptable for old filesystems */
-static inline bool old_valid_dev(dev_t dev)
+static __always_inline bool old_valid_dev(dev_t dev)
 {
 	return MAJOR(dev) < 256 && MINOR(dev) < 256;
 }
 
-static inline u16 old_encode_dev(dev_t dev)
+static __always_inline u16 old_encode_dev(dev_t dev)
 {
 	return (MAJOR(dev) << 8) | MINOR(dev);
 }
 
-static inline dev_t old_decode_dev(u16 val)
+static __always_inline dev_t old_decode_dev(u16 val)
 {
 	return MKDEV((val >> 8) & 255, val & 255);
 }
 
-static inline u32 new_encode_dev(dev_t dev)
+static __always_inline u32 new_encode_dev(dev_t dev)
 {
 	unsigned major = MAJOR(dev);
 	unsigned minor = MINOR(dev);
 	return (minor & 0xff) | (major << 8) | ((minor & ~0xff) << 12);
 }
 
-static inline dev_t new_decode_dev(u32 dev)
+static __always_inline dev_t new_decode_dev(u32 dev)
 {
 	unsigned major = (dev & 0xfff00) >> 8;
 	unsigned minor = (dev & 0xff) | ((dev >> 12) & 0xfff00);
 	return MKDEV(major, minor);
 }
 
-static inline u64 huge_encode_dev(dev_t dev)
+static __always_inline u64 huge_encode_dev(dev_t dev)
 {
 	return new_encode_dev(dev);
 }
 
-static inline dev_t huge_decode_dev(u64 dev)
+static __always_inline dev_t huge_decode_dev(u64 dev)
 {
 	return new_decode_dev(dev);
 }
 
-static inline int sysv_valid_dev(dev_t dev)
+static __always_inline int sysv_valid_dev(dev_t dev)
 {
 	return MAJOR(dev) < (1<<14) && MINOR(dev) < (1<<18);
 }
 
-static inline u32 sysv_encode_dev(dev_t dev)
+static __always_inline u32 sysv_encode_dev(dev_t dev)
 {
 	return MINOR(dev) | (MAJOR(dev) << 18);
 }
 
-static inline unsigned sysv_major(u32 dev)
+static __always_inline unsigned sysv_major(u32 dev)
 {
 	return (dev >> 18) & 0x3fff;
 }
 
-static inline unsigned sysv_minor(u32 dev)
+static __always_inline unsigned sysv_minor(u32 dev)
 {
 	return dev & 0x3ffff;
 }
Signature · 1 line
2.29.2
Acked-by on v1 1/1
#1 lore raw
Quoted 8 lines · Josh Poimboeuf

Silly GCC doesn't always inline these trivial functions.

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

Thanks.

Quoted 3 lines
---
 include/linux/kdev_t.h | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

~Randy

#2 lore raw
In reply to Josh Poimboeuf
Quoted 8 lines · Josh Poimboeuf

Silly GCC doesn't always inline these trivial functions.

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

The alternative is something like:

arch/x86/kernel/sys_ia32.c +7 -2
index 6cf65397d225..6ca3da08dbcb 100644
@@ -133,18 +133,23 @@ static int cp_stat64(struct stat64 __user *ubuf, struct kstat *stat)
 {
 	typeof(ubuf->st_uid) uid = 0;
 	typeof(ubuf->st_gid) gid = 0;
+	u64 dev, rdev;
+
 	SET_UID(uid, from_kuid_munged(current_user_ns(), stat->uid));
 	SET_GID(gid, from_kgid_munged(current_user_ns(), stat->gid));
+	dev = huge_encode_dev(stat->dev);
+	rdev = huge_encode_dev(stat->rdev);
+
 	if (!user_write_access_begin(ubuf, sizeof(struct stat64)))
 		return -EFAULT;
-	unsafe_put_user(huge_encode_dev(stat->dev), &ubuf->st_dev, Efault);
+	unsafe_put_user(dev, &ubuf->st_dev, Efault);
 	unsafe_put_user(stat->ino, &ubuf->__st_ino, Efault);
 	unsafe_put_user(stat->ino, &ubuf->st_ino, Efault);
 	unsafe_put_user(stat->mode, &ubuf->st_mode, Efault);
 	unsafe_put_user(stat->nlink, &ubuf->st_nlink, Efault);
 	unsafe_put_user(uid, &ubuf->st_uid, Efault);
 	unsafe_put_user(gid, &ubuf->st_gid, Efault);
-	unsafe_put_user(huge_encode_dev(stat->rdev), &ubuf->st_rdev, Efault);
+	unsafe_put_user(rdev, &ubuf->st_rdev, Efault);
 	unsafe_put_user(stat->size, &ubuf->st_size, Efault);
 	unsafe_put_user(stat->atime.tv_sec, &ubuf->st_atime, Efault);
 	unsafe_put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec, Efault);
#3 lore raw
Quoted 11 lines · Peter Zijlstra
Quoted 8 lines · Josh Poimboeuf

Silly GCC doesn't always inline these trivial functions.

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

The alternative is something like:

I like your approach better, it avoids __always_inline sprawl.

Peter, care to submit a proper patch?

Andrew, can you drop mine?

Quoted 32 lines
arch/x86/kernel/sys_ia32.c +7 -2
index 6cf65397d225..6ca3da08dbcb 100644
@@ -133,18 +133,23 @@ static int cp_stat64(struct stat64 __user *ubuf, struct kstat *stat)
 {
 	typeof(ubuf->st_uid) uid = 0;
 	typeof(ubuf->st_gid) gid = 0;
+	u64 dev, rdev;
+
 	SET_UID(uid, from_kuid_munged(current_user_ns(), stat->uid));
 	SET_GID(gid, from_kgid_munged(current_user_ns(), stat->gid));
+	dev = huge_encode_dev(stat->dev);
+	rdev = huge_encode_dev(stat->rdev);
+
 	if (!user_write_access_begin(ubuf, sizeof(struct stat64)))
 		return -EFAULT;
-	unsafe_put_user(huge_encode_dev(stat->dev), &ubuf->st_dev, Efault);
+	unsafe_put_user(dev, &ubuf->st_dev, Efault);
 	unsafe_put_user(stat->ino, &ubuf->__st_ino, Efault);
 	unsafe_put_user(stat->ino, &ubuf->st_ino, Efault);
 	unsafe_put_user(stat->mode, &ubuf->st_mode, Efault);
 	unsafe_put_user(stat->nlink, &ubuf->st_nlink, Efault);
 	unsafe_put_user(uid, &ubuf->st_uid, Efault);
 	unsafe_put_user(gid, &ubuf->st_gid, Efault);
-	unsafe_put_user(huge_encode_dev(stat->rdev), &ubuf->st_rdev, Efault);
+	unsafe_put_user(rdev, &ubuf->st_rdev, Efault);
 	unsafe_put_user(stat->size, &ubuf->st_size, Efault);
 	unsafe_put_user(stat->atime.tv_sec, &ubuf->st_atime, Efault);
 	unsafe_put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec, Efault);
Signature · 1 line
Josh
#4 lore raw

PATCH 1/1 x86/compat: Pull huge_encode_dev() outside of UACCESS

Quoted 1 line · Josh Poimboeuf

Peter, care to submit a proper patch?

Here goes..

---
Subject: x86/compat: Pull huge_encode_dev() outside of UACCESS
From: Peter Zijlstra <[email protected]>
Date: Mon, 4 Jan 2021 13:28:25 +0100

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

---
 arch/x86/kernel/sys_ia32.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
arch/x86/kernel/sys_ia32.c +7 -2
@@ -133,18 +133,23 @@ static int cp_stat64(struct stat64 __use
 {
 	typeof(ubuf->st_uid) uid = 0;
 	typeof(ubuf->st_gid) gid = 0;
+	u64 dev, rdev;
+
 	SET_UID(uid, from_kuid_munged(current_user_ns(), stat->uid));
 	SET_GID(gid, from_kgid_munged(current_user_ns(), stat->gid));
+	dev = huge_encode_dev(stat->dev);
+	rdev = huge_encode_dev(stat->rdev);
+
 	if (!user_write_access_begin(ubuf, sizeof(struct stat64)))
 		return -EFAULT;
-	unsafe_put_user(huge_encode_dev(stat->dev), &ubuf->st_dev, Efault);
+	unsafe_put_user(dev, &ubuf->st_dev, Efault);
 	unsafe_put_user(stat->ino, &ubuf->__st_ino, Efault);
 	unsafe_put_user(stat->ino, &ubuf->st_ino, Efault);
 	unsafe_put_user(stat->mode, &ubuf->st_mode, Efault);
 	unsafe_put_user(stat->nlink, &ubuf->st_nlink, Efault);
 	unsafe_put_user(uid, &ubuf->st_uid, Efault);
 	unsafe_put_user(gid, &ubuf->st_gid, Efault);
-	unsafe_put_user(huge_encode_dev(stat->rdev), &ubuf->st_rdev, Efault);
+	unsafe_put_user(rdev, &ubuf->st_rdev, Efault);
 	unsafe_put_user(stat->size, &ubuf->st_size, Efault);
 	unsafe_put_user(stat->atime.tv_sec, &ubuf->st_atime, Efault);
 	unsafe_put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec, Efault);
Acked-by on v1 1/1
#5 lore raw

Re: [PATCH] x86/compat: Pull huge_encode_dev() outside of UACCESS

Quoted 10 lines · Peter Zijlstra
Quoted 1 line · Josh Poimboeuf

Peter, care to submit a proper patch?

Here goes..

---
Subject: x86/compat: Pull huge_encode_dev() outside of UACCESS
From: Peter Zijlstra <[email protected]>
Date: Mon, 4 Jan 2021 13:28:25 +0100

Quoted 6 lines

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

Thanks.

Quoted 33 lines
---
 arch/x86/kernel/sys_ia32.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
arch/x86/kernel/sys_ia32.c +7 -2
@@ -133,18 +133,23 @@ static int cp_stat64(struct stat64 __use
 {
 	typeof(ubuf->st_uid) uid = 0;
 	typeof(ubuf->st_gid) gid = 0;
+	u64 dev, rdev;
+
 	SET_UID(uid, from_kuid_munged(current_user_ns(), stat->uid));
 	SET_GID(gid, from_kgid_munged(current_user_ns(), stat->gid));
+	dev = huge_encode_dev(stat->dev);
+	rdev = huge_encode_dev(stat->rdev);
+
 	if (!user_write_access_begin(ubuf, sizeof(struct stat64)))
 		return -EFAULT;
-	unsafe_put_user(huge_encode_dev(stat->dev), &ubuf->st_dev, Efault);
+	unsafe_put_user(dev, &ubuf->st_dev, Efault);
 	unsafe_put_user(stat->ino, &ubuf->__st_ino, Efault);
 	unsafe_put_user(stat->ino, &ubuf->st_ino, Efault);
 	unsafe_put_user(stat->mode, &ubuf->st_mode, Efault);
 	unsafe_put_user(stat->nlink, &ubuf->st_nlink, Efault);
 	unsafe_put_user(uid, &ubuf->st_uid, Efault);
 	unsafe_put_user(gid, &ubuf->st_gid, Efault);
-	unsafe_put_user(huge_encode_dev(stat->rdev), &ubuf->st_rdev, Efault);
+	unsafe_put_user(rdev, &ubuf->st_rdev, Efault);
 	unsafe_put_user(stat->size, &ubuf->st_size, Efault);
 	unsafe_put_user(stat->atime.tv_sec, &ubuf->st_atime, Efault);
 	unsafe_put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec, Efault);
Signature · 1 line
~Randy
author Acked-by on v1 1/1
#6 lore raw

Re: [PATCH] x86/compat: Pull huge_encode_dev() outside of UACCESS

In reply to Peter Zijlstra
Quoted 10 lines · Peter Zijlstra
Quoted 1 line · Josh Poimboeuf

Peter, care to submit a proper patch?

Here goes..

---
Subject: x86/compat: Pull huge_encode_dev() outside of UACCESS
From: Peter Zijlstra <[email protected]>
Date: Mon, 4 Jan 2021 13:28:25 +0100

Quoted 6 lines

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

Signature · 1 line
Josh
#7 lore raw

Re: [PATCH] x86/compat: Pull huge_encode_dev() outside of UACCESS

Quoted 11 lines · Josh Poimboeuf
Quoted 10 lines · Peter Zijlstra
Quoted 1 line · Josh Poimboeuf

Peter, care to submit a proper patch?

Here goes..

---
Subject: x86/compat: Pull huge_encode_dev() outside of UACCESS
From: Peter Zijlstra <[email protected]>
Date: Mon, 4 Jan 2021 13:28:25 +0100

Quoted 8 lines
Quoted 6 lines

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

Actually the other patch was already merged into Linus' tree.

Maybe add a revert of the other patch to this one?

Signature · 1 line
Josh
Acked-by
#8 lore raw

Re: [PATCH] x86/compat: Pull huge_encode_dev() outside of UACCESS

Quoted 16 lines · Josh Poimboeuf
Quoted 15 lines · Josh Poimboeuf
Quoted 14 lines · Peter Zijlstra
Quoted 1 line · Josh Poimboeuf

Peter, care to submit a proper patch?

Here goes..

---
Subject: x86/compat: Pull huge_encode_dev() outside of UACCESS
From: Peter Zijlstra <[email protected]>
Date: Mon, 4 Jan 2021 13:28:25 +0100

Fixes the following warning:

arch/x86/kernel/sys_ia32.o: warning: objtool: cp_stat64()+0xd8: call to new_encode_dev() with UACCESS enabled

Quoted 9 lines
Quoted 4 lines
Quoted 2 lines

Actually the other patch was already merged into Linus' tree.

Maybe add a revert of the other patch to this one?

Glad you noticed.

I reverted the kdev_t patch and applied Peter's patch and it still tested OK, so (still):

Signature · 1 line
~Randy