From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1) Description of problem: When executing pppd to connect to a remote server we get the following error: Couldn't get channel number: Bad address Afer further investigation and research on the internet I found there is a problem with the ppp_async module and that it could be fixed by replacing put_user with copy_to_user in a few of the ioctl calls. I have made the modifications to the code thanks to a patch sent to me by an unknown source and it fixes the problem. Unfortunately, asking our customers to rebuild the modules in the kernel is not an acceptable solution. Version-Release number of selected component (if applicable): ppp-2.4.3-ia64 How reproducible: Always Steps to Reproduce: 1.Make a modem connection to a remote server 2.run: pppd /dev/ttyS1 115200 -detach ....... 3. Actual Results: Couldn't get channel number: bad address Expected Results: pppd to return without error and a new ppp0 entry to show up when typing ifconfig Additional info:
Do you had to modify kernel modules?
Brian, could you please attach the patch that you reference. thanks.
Brian, can we get a pointer to that patch? thanks.
http://marc.theaimsgroup.com/?l=linux-ia64&m=105590693005655&w=2
Created attachment 117674 [details] patch to correct use of put_user in ppp.o module Think I've found the problem. Its described pretty accurately here: http://www.gelato.unsw.edu.au/archives/linux-ia64/0312/7755.html The problem appears to be that registers get inappropriately clobbered when functions are inlined as arguments to put_user (specifically to __put_user_check). The solution seems to be to make local copies of the passed in arguments so as to avoid the register clobbering. The fix, as proposed works well in my test case. Given the relative maturity of RHEL3 however, I'd be nervous about proposing a modification to such a widely used macro w/o a lot more testing. Given that, I think the easier solution for the problem at hand would be to simply move the inlined functions out of the put_user/get_user macros in the ppp module, as this patch does. It also passes my test case, and I'm waiting for confirmation from the reporter now.
Created attachment 118056 [details] patch with correct fix for put_user on ia64 Consensus on review was that this should probably be fixed the right way. Fixing __put_user_check so that passing in functions as the value to copy worked, outweighed the potential risk of modifiying such a pervasive function. This patch corrects __put_user_check so that passed in functions are evaluated prior to being passed to __put_user_size.
A fix for this problem has just been committed to the RHEL3 U7 patch pool this evening (in kernel version 2.4.21-37.2.EL).
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2006-0144.html