Re: Fix for __div64_32 locks when using some 64 bit numbers
From: davidastro <hidden>
Date: 2009-03-20 19:33:35
Hi Ben: I was wondering if you have any change to look into and test the propose fix I suggested in my previous post. I'd like to know if the fix is correct. Thanks for your attention, Benjamin Herrenschmidt wrote:
On Tue, 2009-03-17 at 14:15 -0700, davidastro wrote:quoted
I found a bug when using the function __div64_32 in assembly in a 32 bit ppc architecture unit. I tried the numbers 55834565048000000 for the dividend and 4294967079 for the divisor. When passing these two numbers to the function __div64_32, I had a software lock. I searched for possible patches online and in different forums but I could not find anything related to the assembly implementation to this function (I would have to apologize if somebody already found a fix :-) ). Anyway, when analyzing the assembly code, I found out with gdb the problem. I am not an expert in ppc architecture but I read the documentation and I am pretty sure I solved the issue (I have been testing for couple of days using random 64 to 32 number combinations with good results). Who or Where should I post the fix to be reviewed.Here is fine :-) Ben. _______________________________________________ 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-tp22567864p22627440.html Sent from the linuxppc-dev mailing list archive at Nabble.com.