arm64: fix missing __user in compat_vfp_{get,set}()
Subsystems (2)
PATCH 1/1 arm64: fix missing __user in compat_vfp_{get,set}()
compat_vfp_get() & compat_vfp_set() are two helpers reading or writting some values via {get,put}_user() which need a pointer annotated with '__user'.
The buffers used by the two helpers are correctly '__user' annotated but need to be casted to a real type before being given to {get,put}_user().
The problem is that this cast lack a '__user' annotation.
Fix this by adding the missing '__user'.
--- arch/arm64/kernel/ptrace.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/kernel/ptrace.c +2 -2
@@ -908,7 +908,7 @@ static int compat_vfp_get(struct task_struct *target,if (count && !ret) { fpscr = (uregs->fpsr &VFP_FPSCR_STAT_MASK ) | (uregs->fpcr &VFP_FPSCR_CTRL_MASK ); - ret =put_user (fpscr, (compat_ulong_t *)ubuf); + ret =put_user (fpscr, (compat_ulong_t __user *)ubuf); }return ret;
@@ -932,7 +932,7 @@ static int compat_vfp_set(struct task_struct *target, VFP_STATE_SIZE - sizeof(compat_ulong_t));if (count && !ret) { - ret =get_user (fpscr, (compat_ulong_t *)ubuf); + ret =get_user (fpscr, (compat_ulong_t __user *)ubuf); uregs->fpsr = fpscr &VFP_FPSCR_STAT_MASK ; uregs->fpcr = fpscr &VFP_FPSCR_CTRL_MASK ; }
Signature · 1 line
2.13.0
PATCH 1/1 arm64: fix missing __user in compat_vfp_{get,set}()
Quoted 10 lines · Luc Van Oostenryck
compat_vfp_get()&compat_vfp_set()are two helpers reading or writting some values via {get,put}_user() which need a pointer annotated with '__user'.
The buffers used by the two helpers are correctly '__user' annotated but need to be casted to a real type before being given to {get,put}_user().The problem is that this cast lack a '
__user' annotation.Fix this by adding the missing '
__user'.
Casting __user off is something that sparse certainly should be warning about (and it was, by the looks of it).
(There remains a sparse warning in compat_ptrace_write_user(), where we pass implicitly cast __user onto a pointer, under set_fs(KERNEL_DS). I'm not sure what the most correct fix is there.)
Cheers
---Dave
Quoted 36 lines
Signed-off-by Luc Van Oostenryck <[email protected]>--- arch/arm64/kernel/ptrace.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)arch/arm64/kernel/ptrace.c +2
-2@@ -908,7 +908,7 @@ static int compat_vfp_get(struct task_struct *target,if (count && !ret) { fpscr = (uregs->fpsr &VFP_FPSCR_STAT_MASK ) | (uregs->fpcr &VFP_FPSCR_CTRL_MASK ); - ret =put_user (fpscr, (compat_ulong_t *)ubuf); + ret =put_user (fpscr, (compat_ulong_t __user *)ubuf); }return ret;@@ -932,7 +932,7 @@ static int compat_vfp_set(struct task_struct *target, VFP_STATE_SIZE - sizeof(compat_ulong_t));if (count && !ret) { - ret =get_user (fpscr, (compat_ulong_t *)ubuf); + ret =get_user (fpscr, (compat_ulong_t __user *)ubuf); uregs->fpsr = fpscr &VFP_FPSCR_STAT_MASK ; uregs->fpcr = fpscr &VFP_FPSCR_CTRL_MASK ; }Signature · 7 lines
2.13.0 _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 1/1 arm64: fix missing __user in compat_vfp_{get,set}()
Quoted 3 lines · Dave Martin
(There remains a sparse warning in
compat_ptrace_write_user(), where we pass implicitly cast__useronto a pointer, underset_fs(KERNEL_DS). I'm not sure what the most correct fix is there.)
Yes, I saw.
There is no pretty solution for things like here:
you can't adjust the type, you're forced to make a cast, explicit or implicit.
Regards,
-- Luc
PATCH 1/1 arm64: fix missing __user in compat_vfp_{get,set}()
In reply to Luc Van OostenryckQuoted 7 lines · Luc Van Oostenryck
compat_vfp_get()&compat_vfp_set()are two helpers reading or writting some values via {get,put}_user() which need a pointer annotated with '__user'.
The buffers used by the two helpers are correctly '__user' annotated but need to be casted to a real type before being given to {get,put}_user().
Quoted 8 lines
The problem is that this cast lack a '
__user' annotation.Fix this by adding the missing '
__user'.Signed-off-by Luc Van Oostenryck <[email protected]>--- arch/arm64/kernel/ptrace.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
This doesn't apply against the arm64 for-next/core branch, due to changes from Dave reworking this to use user_regset_copyout.
Will
PATCH v2 1/1 arm64: fix missing __user in compat_vfp_set()
compat_vfp_set() is a helper writting some values via put_user() and put_user() need a pointer annotated with '__user'. The buffer used by the helper is correctly annotated with '__user' but need to be casted to a real type before being given to put_user().
The problem is that this cast lack a '__user' annotation.
Fix this by adding the missing '__user'.
---
Change since v1:
- rebase against arm64/for-next/core
- drop the change for compat_vfp_get() which is no more needed
--- arch/arm64/kernel/ptrace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/kernel/ptrace.c +1 -1
@@ -946,7 +946,7 @@ static int compat_vfp_set(struct task_struct *target, VFP_STATE_SIZE - sizeof(compat_ulong_t));if (count && !ret) { - ret =get_user (fpscr, (compat_ulong_t *)ubuf); + ret =get_user (fpscr, (compat_ulong_t __user *)ubuf); uregs->fpsr = fpscr &VFP_FPSCR_STAT_MASK ; uregs->fpcr = fpscr &VFP_FPSCR_CTRL_MASK ; }
Signature · 1 line
2.13.0
PATCH v2 1/1 arm64: fix missing __user in compat_vfp_set()
Quoted 11 lines · Luc Van Oostenryck
compat_vfp_set()is a helper writting some values viaput_user()andput_user()need a pointer annotated with '__user'. The buffer used by the helper is correctly annotated with '__user' but need to be casted to a real type before being given toput_user().The problem is that this cast lack a '
__user' annotation.Fix this by adding the missing '
__user'.Signed-off-by Luc Van Oostenryck <[email protected]>
Apologies, I has half a reply written and then realised this issue was less straightforward that I first thought. I have an alternate patch, see [1].
Cheers
---Dave
[1] [PATCH 0/3] Miscellaneous minor compat ptrace fixes
http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/516592.html
Quoted 31 lines
---
Change since v1:
- rebase against arm64/for-next/core
- drop the change forcompat_vfp_get()which is no more needed--- arch/arm64/kernel/ptrace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/kernel/ptrace.c +1
-1@@ -946,7 +946,7 @@ static int compat_vfp_set(struct task_struct *target, VFP_STATE_SIZE - sizeof(compat_ulong_t));if (count && !ret) { - ret =get_user (fpscr, (compat_ulong_t *)ubuf); + ret =get_user (fpscr, (compat_ulong_t __user *)ubuf); uregs->fpsr = fpscr &VFP_FPSCR_STAT_MASK ; uregs->fpcr = fpscr &VFP_FPSCR_CTRL_MASK ; }Signature · 8 lines
2.13.0 _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH v2 1/1 arm64: fix missing __user in compat_vfp_set()
Quoted 9 lines · Dave Martin
Quoted 8 lines · Luc Van Oostenryck
compat_vfp_set()is a helper writting some values viaput_user()andput_user()need a pointer annotated with '__user'. The buffer used by the helper is correctly annotated with '__user' but need to be casted to a real type before being given toput_user().The problem is that this cast lack a '
__user' annotation.
Quoted 7 lines
Quoted 3 lines
Fix this by adding the missing '
__user'.Signed-off-by Luc Van Oostenryck <[email protected]>Apologies, I has half a reply written and then realised this issue was less straightforward that I first thought. I have an alternate patch, see [1].
Ah yes.
I saw the now asymmetry between the set/copyout and the set/copyin and I was a bit surprised by it.
Your new version look better (but I can't judge more than the look).
-- Luc