Re: [linuxsh-dev] /bin/mount crashes on 2.4.0-test1...

[NIIBE Y. <gn...@ch...>]

Stuart Menefy wrote:
 > Yes. The problem appears to be caused by a fix which was put into the
 > generic mount code (in copy_mount_options). This now correctly checks
 > the address the user gave for the file system type. However this has
 > highlighted a bug in the SuperH code. __copy_user should return the
 > number of bytes not copied (ie 0 for a success). However the SH version
 > was returning EFAULT for a failure. The attached patch should fix
 > this.
[...]
 > Could somebody check my resoning? If its sound I'll check in, together
 > with the other changes.

I've checked.  Your analysis is right.  It should not return -EFAULT, 
it's put_user/get_user which return -EFAULT, there was some confusion.

 > ! 		"	.long 9b,2b\n"

This is right, the return value is remaining bytes.

 > ! 		" add	#1, %0\n\t"
[...]
 > ! 		"	.long 1b,3b\n"

This fix is also right.  We should add +1 to get the value.

Thank you.
--