parisc: Fix user access miscompilation
Helge Deller
maintainerPATCH 1/1 parisc: Fix user access miscompilation
After commit 4b9d2a731c3d ("parisc: Switch user access functions to signal errors in r29 instead of r8") bash suddenly started to report those warnings after login:
-bash: cannot set terminal process group (-1): Bad file descriptor -bash: no job control in this shell
It turned out, that another function call inside a put_user(), e.g.:
put_user(vt_do_kdgkbmode(console), (int __user *)arg);
clobbered the error register (r29) and thus failed.
Avoid this miscompilation by first calculate the value in __put_user() before calling __put_user_internal() to actually write the value to user space.
Additionally, prefer the "+" constraint on pu_err and gu_err registers to tell the compiler that those operands are both read and written by the assembly instruction.
arch/parisc/include/asm/uaccess.h +8 -7
@@ -89,7 +89,7 @@ struct exception_table_entry { __asm__("1: " ldx" 0(" sr"%2),%0\n" \"9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ - : "=r"(__gu_val), "=r"(__gu_err) \ + : "=&r"(__gu_val), "+r"(__gu_err) \ : "r"(ptr), "1"(__gu_err)); \\ (val) = (__force __typeof__(*(ptr))) __gu_val; \
@@ -123,7 +123,7 @@ struct exception_table_entry {"9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ASM_EXCEPTIONTABLE_ENTRY_EFAULT (2b ,9b ) \ - : "=&r"(__gu_tmp.l), "=r"(__gu_err) \ + : "=&r"(__gu_tmp.l), "+r"(__gu_err) \ : "r"(ptr), "1"(__gu_err)); \\ (val) = __gu_tmp.t; \
@@ -132,10 +132,9 @@ struct exception_table_entry {#endif /* !defined(CONFIG_64BIT) */ -#define __put_user_internal (sr, x, ptr) \ +#define __put_user_internal (sr, __x, ptr) \ ({ \ ASM_EXCEPTIONTABLE_VAR(__pu_err); \ - __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x); \ \ switch (sizeof(*(ptr))) { \ case 1: __put_user_asm(sr, "stb", __x, ptr); break; \
@@ -150,7 +149,9 @@ struct exception_table_entry {#define __put_user (x, ptr) \ ({ \ - __put_user_internal("%%sr3,", x, ptr); \ + __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x); \ + \ + __put_user_internal("%%sr3,", __x, ptr); \ })#define __put_kernel_nofault (dst, src, type, err_label) \
@@ -180,7 +181,7 @@ struct exception_table_entry {"1: " stx" %2,0(" sr"%1)\n" \"9:\n" \ASM_EXCEPTIONTABLE_ENTRY_EFAULT (1b ,9b ) \ - : "=r"(__pu_err) \ + : "+r"(__pu_err) \ : "r"(ptr), "r"(x), "0"(__pu_err))
@@ -193,7 +194,7 @@ struct exception_table_entry {"9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ASM_EXCEPTIONTABLE_ENTRY_EFAULT (2b ,9b ) \ - : "=r"(__pu_err) \ + : "+r"(__pu_err) \ : "r"(ptr), "r"(__val), "0"(__pu_err)); \ } while (0)
Quoted 7 lines · Helge Deller
After commit
4b9d2a731c3d("parisc: Switch user access functions to signal errors in r29 instead of r8") bash suddenly started to report those warnings after login:(quoted diff) +0
-2-bash: cannot set terminal process group (-1): Bad file descriptor -bash: no job control in this shell
Quoted 6 lines
It turned out, that another function call inside a
put_user(), e.g.:put_user(vt_do_kdgkbmode(console), (int __user *)arg);clobbered the error register (r29) and thus failed. Avoid this miscompilation by first calculate the value in
__put_user()before calling__put_user_internal()to actually write the value to user space.
Couldn't the same issue occur with ptr argument in these macros?
Quoted 13 lines
Additionally, prefer the "+" constraint on
pu_errandgu_errregisters to tell the compiler that those operands are both read and written by the assembly instruction.Signed-off-by Helge Deller <[email protected]>Fixes 4b9d2a731c3d ("parisc: Switch user access functions to signal errors in r29 instead of r8")arch/parisc/include/asm/uaccess.h +0
-0
arch/parisc/include/asm/uaccess.h struct exception_table_entry { · line 92
Quoted 5 lines
arch/parisc/include/asm/uaccess.h +1
-1@@ -89,7 +89,7 @@ struct exception_table_entry { __asm__("1: " ldx" 0(" sr"%2),%0\n" \"9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ - : "=r"(__gu_val), "=r"(__gu_err) \ + : "=&r"(__gu_val), "+r"(__gu_err) \
I don't believe the early clobber is needed on __gnu_val.
Quoted 25 lines
arch/parisc/include/asm/uaccess.h +2
-3: "r"(ptr), "1"(__gu_err)); \ \ (val) = (__force __typeof__(*(ptr))) __gu_val; \@@ -123,7 +123,7 @@ struct exception_table_entry {"9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ASM_EXCEPTIONTABLE_ENTRY_EFAULT (2b ,9b ) \ - : "=&r"(__gu_tmp.l), "=r"(__gu_err) \ + : "=&r"(__gu_tmp.l), "+r"(__gu_err) \ : "r"(ptr), "1"(__gu_err)); \\ (val) = __gu_tmp.t; \ @@ -132,10 +132,9 @@ struct exception_table_entry {#endif /* !defined(CONFIG_64BIT) */ -#define __put_user_internal (sr, x, ptr) \ +#define __put_user_internal (sr, __x, ptr) \ ({ \ ASM_EXCEPTIONTABLE_VAR(__pu_err); \ - __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x); \ \ switch (sizeof(*(ptr))) { \ case 1: __put_user_asm(sr, "stb", __x, ptr); break; \
arch/parisc/include/asm/uaccess.h struct exception_table_entry { · line 152
Quoted 5 lines
arch/parisc/include/asm/uaccess.h +1
-1@@ -150,7 +149,9 @@ struct exception_table_entry {#define __put_user (x, ptr) \ ({ \ - __put_user_internal("%%sr3,", x, ptr); \ + __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x); \
Whitespace?
Quoted 15 lines
arch/parisc/include/asm/uaccess.h +3
-1+ \ +__put_user_internal ("%%sr3," , __x, ptr); \ })#define __put_kernel_nofault (dst, src, type, err_label) \@@ -180,7 +181,7 @@ struct exception_table_entry {"1: " stx" %2,0(" sr"%1)\n" \"9:\n" \ASM_EXCEPTIONTABLE_ENTRY_EFAULT (1b ,9b ) \ - : "=r"(__pu_err) \ + : "+r"(__pu_err) \ : "r"(ptr), "r"(x), "0"(__pu_err))
arch/parisc/include/asm/uaccess.h struct exception_table_entry { · line 200
Quoted 8 lines
arch/parisc/include/asm/uaccess.h +1
-1@@ -193,7 +194,7 @@ struct exception_table_entry {"9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ASM_EXCEPTIONTABLE_ENTRY_EFAULT (2b ,9b ) \ - : "=r"(__pu_err) \ + : "+r"(__pu_err) \ : "r"(ptr), "r"(__val), "0"(__pu_err)); \ } while (0)
Dave
Signature · 1 line
John David Anglin [email protected]
Helge Deller
maintainerQuoted 8 lines · John David Anglin
Quoted 7 lines · Helge Deller
After commit
4b9d2a731c3d("parisc: Switch user access functions to signal errors in r29 instead of r8") bash suddenly started to report those warnings after login:(quoted diff) +0
-2-bash: cannot set terminal process group (-1): Bad file descriptor -bash: no job control in this shell
Quoted 7 lines
Quoted 6 lines
It turned out, that another function call inside a
put_user(), e.g.:
put_user(vt_do_kdgkbmode(console), (int __user *)arg); clobbered the error register (r29) and thus failed.
Avoid this miscompilation by first calculate the value in__put_user()before calling__put_user_internal()to actually write the value to user space.Couldn't the same issue occur with ptr argument in these macros?
Theoretically, yes. Haven't seen it though.
Quoted 7 lines
Quoted 7 lines
Additionally, prefer the "+" constraint on
pu_errandgu_errregisters to tell the compiler that those operands are both read and written by the assembly instruction.Signed-off-by Helge Deller <[email protected]>Fixes 4b9d2a731c3d ("parisc: Switch user access functions to signal errors in r29 instead of r8")
Quoted 11 lines
arch/parisc/include/asm/uaccess.h struct exception_table_entry { · line 92
Quoted 10 lines
arch/parisc/include/asm/uaccess.h +1
-1@@ -89,7 +89,7 @@ struct exception_table_entry { __asm__("1: " ldx " 0(" sr "%2),%0\n" \ "9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ - : "=r"(__gu_val), "=r"(__gu_err) \ + : "=&r"(__gu_val), "+r"(__gu_err) \I don't believe the early clobber is needed on
__gnu_val.
Right.
Quoted 26 lines
Quoted 26 lines
arch/parisc/include/asm/uaccess.h +2
-3: "r"(ptr), "1"(__gu_err)); \ \ (val) = (__force __typeof__(*(ptr))) __gu_val; \@@ -123,7 +123,7 @@ struct exception_table_entry { "9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(2b, 9b) \ - : "=&r"(__gu_tmp.l), "=r"(__gu_err) \ + : "=&r"(__gu_tmp.l), "+r"(__gu_err) \ : "r"(ptr), "1"(__gu_err)); \ \ (val) = __gu_tmp.t; \@@ -132,10 +132,9 @@ struct exception_table_entry { #endif /* !defined(CONFIG_64BIT) */ -#define __put_user_internal (sr, x, ptr) \ +#define __put_user_internal (sr, __x, ptr) \ ({ \ ASM_EXCEPTIONTABLE_VAR(__pu_err); \ - __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x); \ \ switch (sizeof(*(ptr))) { \ case 1: __put_user_asm(sr, "stb", __x, ptr); break; \@@ -150,7 +149,9 @@ struct exception_table_entry {
Quoted 5 lines
arch/parisc/include/asm/uaccess.h struct exception_table_entry { · line 152
Quoted 4 lines
arch/parisc/include/asm/uaccess.h +1
-1#define __put_user(x, ptr) \ ({ \ - __put_user_internal("%%sr3,", x, ptr); \ + __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x); \Whitespace?
Will fix.
Quoted 14 lines
Quoted 14 lines
arch/parisc/include/asm/uaccess.h +3
-1+ \ + __put_user_internal("%%sr3,", __x, ptr); \ }) #define __put_kernel_nofault(dst, src, type, err_label) \@@ -180,7 +181,7 @@ struct exception_table_entry { "1: " stx " %2,0(" sr "%1)\n" \ "9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ - : "=r"(__pu_err) \ + : "+r"(__pu_err) \ : "r"(ptr), "r"(x), "0"(__pu_err))
Quoted 8 lines
arch/parisc/include/asm/uaccess.h struct exception_table_entry { · line 199
Quoted 8 lines
arch/parisc/include/asm/uaccess.h +1
-1@@ -193,7 +194,7 @@ struct exception_table_entry { "9:\n" \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 9b) \ ASM_EXCEPTIONTABLE_ENTRY_EFAULT(2b, 9b) \ - : "=r"(__pu_err) \ + : "+r"(__pu_err) \ : "r"(ptr), "r"(__val), "0"(__pu_err)); \ } while (0)
Overall, this current patch turned out to not cover all cases. I think it's too complicated to really try prevent the compiler to optimize it here, so I'll send out a new patch shortly which simply initializes __gu_err and __pu_err in the assembly statement itself. My other codings to avoid that would increase the code a lot, and I'm not sure if it's even then possible to avoid the failure always.
Helge