Re: Fix for __div64_32 locks when using some 64 bit numbers

Paul,

You are correct. Since it was such a small piece of code, I didn't use diff
between the kernel I am using (2.6.14) and the newest ones (2.6.28 on the
Linux). Also, I do not know how to look in the kernel files commit history.
I guess the link you gave me might be of some help. 

Thanks for your help


Paul Mackerras wrote:

davidastro [off-list ref] writes:

quoted

Basically, the numbers shown above was causing the 64 by 32 bit algorithm
to
divide by zero making the unit spin and also giving incorrect results.
Here is the code as it was before.

[snip]

quoted

        cntlzw  r0,r5           # we are shifting the dividend right
        li      r10,-1          # to make it < 2^32, and shifting
        srw     r10,r10,r0      # the divisor right the same amount,
        add     r9,r4,r10       # rounding up (so the estimate cannot
        andc    r11,r6,r10      # ever be too large, only too small)
        andc    r9,r9,r10      #THIS CODE COULD STORE A ZERO IN r9
        or      r11,r5,r11
        rotlw   r9,r9,r0
        rotlw   r11,r11,r0

That bug was fixed in October 2005 in commit 344480b99730bfd2, and the
fix is in v2.6.15-rc1 and all later kernels.  I fixed it a bit
differently to your suggestion though - see

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=344480b99730bfd205e306d3fd168cdcebe83425

You must be working from an old kernel tree - which kernel version are
you using?

Paul.
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev

-- 
View this message in context: http://www.nabble.com/Fix-for-__div64_32-locks-when-using-some-64-bit-numbers-tp22567864p22665081.html
Sent from the linuxppc-dev mailing list archive at Nabble.com.