Skip to contentSkip to conversation

fix long long cast in pte macro

RFC PATCH In review Willy Tarreau 2 participants 3 messages Aug 30, 2006 – Aug 31, 2006

lore
Share & tools
lore

Willy Tarreau

author
#0 lore raw

RFC PATCH 1/1 fix long long cast in pte macro

Hi,

PaX Team sent me this patch, which I think is valid. It fixes a long long cast in the pte macro for i386 and mips. If nobody has any objections, I will apply it to 2.4. I'd also like someone to check whether it's needed for 2.6 and to forward port it if needed.

Thanks,
Willy

--

the current idiom used for initializing a structure of two unsigned longs from unsigned long long is wrong, it effectively loses the upper 32 bits which in this particular case could turn a non-executable PTE into an executable one on NX capable i386 (i.e., it's a potential security bug). fortunately the in-tree users in 2.4 (drivers/char/drm-4.0/ffb_drv.c and arch/mips/baget/baget.c) are not affected.

include/asm-i386/page.h +2 -1
diff -u linux-2.4.33-pax/include/asm-i386/page.h linux-2.4.33-pax/include/asm-i386/page.h
@@ -41,11 +41,13 @@
 typedef struct { unsigned long long pmd; } pmd_t;
 typedef struct { unsigned long long pgd; } pgd_t;
 #define pte_val(x)	((x).pte_low | ((unsigned long long)(x).pte_high << 32))
+#define __pte(x) ({ pte_t __pte = {(x), (x) >> 32}; __pte; })
 #else
 typedef struct { unsigned long pte_low; } pte_t;
 typedef struct { unsigned long pmd; } pmd_t;
 typedef struct { unsigned long pgd; } pgd_t;
 #define pte_val(x)	((x).pte_low)
+#define __pte(x) ((pte_t) { (x) } )
 #endif
 #define PTE_MASK	PAGE_MASK
 

@@ -55,7 +57,6 @@
 #define pgd_val(x)	((x).pgd)
 #define pgprot_val(x)	((x).pgprot)
 
-#define __pte(x) ((pte_t) { (x) } )
 #define __pmd(x) ((pmd_t) { (x) } )
 #define __pgd(x) ((pgd_t) { (x) } )
 #define __pgprot(x)	((pgprot_t) { (x) } )
include/asm-mips/page.h +3 -1
diff -u linux-2.4.33-pax/include/asm-mips/page.h linux-2.4.33-pax/include/asm-mips/page.h
@@ -77,13 +77,16 @@
   #ifdef CONFIG_CPU_MIPS32
     typedef struct { unsigned long pte_low, pte_high; } pte_t;
     #define pte_val(x)    ((x).pte_low | ((unsigned long long)(x).pte_high << 32))
+    #define __pte(x)	({ pte_t __pte = {(x), (x) >> 32}; __pte; })
   #else
     typedef struct { unsigned long long pte_low; } pte_t;
     #define pte_val(x)    ((x).pte_low)
+    #define __pte(x)	((pte_t) { (x) } )
   #endif
 #else
 typedef struct { unsigned long pte_low; } pte_t;
 #define pte_val(x)    ((x).pte_low)
+#define __pte(x)	((pte_t) { (x) } )
 #endif
 
 typedef struct { unsigned long pmd; } pmd_t;

@@ -96,7 +99,6 @@
 
 #define ptep_buddy(x)	((pte_t *)((unsigned long)(x) ^ sizeof(pte_t)))
 
-#define __pte(x)	((pte_t) { (x) } )
 #define __pmd(x)	((pmd_t) { (x) } )
 #define __pgd(x)	((pgd_t) { (x) } )
 #define __pgprot(x)	((pgprot_t) { (x) } )
Acked-by on v1 1/1
#1 lore raw
Quoted 4 lines · Willy Tarreau

PaX Team sent me this patch, which I think is valid. It fixes a long long cast in the pte macro for i386 and mips. If nobody has any objections, I will apply it to 2.4. I'd also like someone to check whether it's needed for 2.6 and to forward port it if needed.

Yes, a similar 2.6 patch is needed as well, I'll care of that. For the MIPS segment of your 2.4 patch:

Acked-by Ralf Baechle <[email protected]>

Ralf

author
#2 lore raw
Quoted 7 lines · Ralf Baechle
Quoted 4 lines · Willy Tarreau

PaX Team sent me this patch, which I think is valid. It fixes a long long cast in the pte macro for i386 and mips. If nobody has any objections, I will apply it to 2.4. I'd also like someone to check whether it's needed for 2.6 and to forward port it if needed.

Quoted 6 lines

Yes, a similar 2.6 patch is needed as well, I'll care of that. For the MIPS segment of your 2.4 patch:

Acked-by Ralf Baechle <[email protected]>

Ralf

Thanks Ralf,
Willy