SELinux & ia64
SELinux appears to cause warnings while attempting mounts both at boot-time and run-time:
kernel unaligned access to 0xa0000002003a0056, ip=0xa0000001002076d0 kernel unaligned access to 0xa0000002003a005e, ip=0xa0000001002076d0 kernel unaligned access to 0xa0000002003a0066, ip=0xa0000001002076d0 kernel unaligned access to 0xa0000002003a006e, ip=0xa0000001002076d0
I've tracked this to the usage of le32_to_cpu in
security/selinux/ss/policydb.c
The code in question uses:
len = le32_to_cpu(buf[0]);
and should be
len = le32_to_cpu(get_unaligned(&buf[0]));
However, this is probably not a good solution as the get_unaligned macro can be expensive
on platforms other than ia64.
I'm tempted to redefine the le32_to_cpu function for policydb.c for the ia64 platform, but
before I go down that road I was wondering if anyone had hit this issue elsewhere in the kernel?
Prarit.
Quoted 11 lines · Prarit Bhargava
I've tracked this to the usage of
le32_to_cpuin
security/selinux/ss/policydb.cThe code in question uses:
len = le32_to_cpu(buf[0]);and should be
len = le32_to_cpu(get_unaligned(&buf[0]));
It would probably be better to find out why buf is unaligned, and see if we can make other changes to make it aligned. If we can't, then I recommend a new macro for accessing this element, rather than nest the macros like this. It could look something like:
len = get_unaligned_le32(buf + 0);
and be implemented perhaps as:
static u32 get_unaligned_le32(u32 *buf)
{
char *bufc = buf;
return *bufc | (*(bufc+1) << 8) | (*(bufc+2) << 16) | (*(bufc+3) << 24);
}Signature · 6 lines
"Next the statesmen will invent cheap lies, putting the blame upon the nation that is attacked, and every man will be glad of those conscience-soothing falsities, and will diligently study them, and refuse to examine any refutations of them; and thus he will by and by convince himself that the war is just, and will thank God for the better sleep he enjoys after this process of grotesque self-deception." -- Mark Twain
Hi Matthew,
Thanks for the reply. In this instance, buf[0] points to a char* -- it might not be aligned to u16.
Prarit
Quoted 34 lines · Matthew Wilcox
Quoted 4 lines · Prarit Bhargava
I've tracked this to the usage of
le32_to_cpuin
security/selinux/ss/policydb.cQuoted 9 lines
The code in question uses:
len = le32_to_cpu(buf[0]);and should be
len = le32_to_cpu(get_unaligned(&buf[0]));It would probably be better to find out why buf is unaligned, and see if we can make other changes to make it aligned. If we can't, then I recommend a new macro for accessing this element, rather than nest the macros like this. It could look something like:
len = get_unaligned_le32(buf + 0);and be implemented perhaps as:
static u32
get_unaligned_le32(u32 *buf){ char *bufc = buf; return *bufc | (*(bufc+1) << 8) | (*(bufc+2) << 16) | (*(bufc+3) << 24); }
This is being discussed further on selinux -- sorry for the crosstalk.
P.
Quoted 34 lines · Matthew Wilcox
Quoted 4 lines · Prarit Bhargava
I've tracked this to the usage of
le32_to_cpuin
security/selinux/ss/policydb.cQuoted 9 lines
The code in question uses:
len = le32_to_cpu(buf[0]);and should be
len = le32_to_cpu(get_unaligned(&buf[0]));It would probably be better to find out why buf is unaligned, and see if we can make other changes to make it aligned. If we can't, then I recommend a new macro for accessing this element, rather than nest the macros like this. It could look something like:
len = get_unaligned_le32(buf + 0);and be implemented perhaps as:
static u32
get_unaligned_le32(u32 *buf){ char *bufc = buf; return *bufc | (*(bufc+1) << 8) | (*(bufc+2) << 16) | (*(bufc+3) << 24); }