Description of problem: Cannot load HTTP websites using an Intellinet 150N wireless adapter since upgrading kernel form 3.4 to 3.5. Version-Release number of selected component (if applicable): Kernel 3.5 Works fine with 3.4.6-2 Fails both 3.5.0-2 AND 3.5.1-1 How reproducible: Try to visit an HTTPS site with Firefox while using an Intellinet 150N wireless adapter Actual results: Error page saying "ssl_error_bad_mac_read" Expected results: It should just work Additional info: N/A
Support for this adapter was added by bug 699618. Also, going back to my older, slower, USB wireless adapter works fine with the 3.5 kernel.
rtl8712 regression :-(
This bug is the same as https://bugzilla.kernel.org/show_bug.cgi?id=45071. This problem was bisected to commit c862815, which is entitled "tcp: reduce out_of_order memory use". As other network drivers are OK with this code, it is obvious that the patch exposed a bug in r8712u. I posted the finding on LKML and to the authors of the patch asking for ideas. but got no response from anyone. I have no idea why the SSL packets are out of order.
Created attachment 606391 [details] This patch reduces the maximum size of a buffer from a ridiculous 30K to 9100 bytes Sorry it took so long to get back to this problem. Difficulty with a different driver also bisected to the same commit. This time, my mail request got some suggestions, which resulted in this trial patch. Please test.
(In reply to comment #4) > Created attachment 606391 [details] > This patch reduces the maximum size of a buffer from a ridiculous 30K to > 9100 bytes > > Sorry it took so long to get back to this problem. Difficulty with a > different driver also bisected to the same commit. This time, my mail > request got some suggestions, which resulted in this trial patch. > > Please test. can this possibly be built and uploaded as a staging update? I really do not have the capability to do anything with this patch without a lot more info on how to build including this patch (which is maybe something I should know how to do)
As I don't use Fedora, it would not be possible for me to build a kernel for you. Perhaps one of the Fedora people (perhaps John Linville, or whomever reads kernel-maint) could incorporate this patch into the Fedora kernel and build a 3.5 version with the patch.
The patch is not correct. Do not bother building a test kernel.
Created attachment 611536 [details] New patch to fix the SSL problem on 64-bit systems This patch from Eric Dumazet fixes the problem for me. A few systems have had problems with the latest firmware, thus a request has been posted to revert the file. This problem affects the stability of the connection, not the SSL difficulty.
I no longer see this issue. Was this fixed upstream?
Yes with commit abf02cfc179bb4bd30d05f582d61b3b8f429b813, which has been backported to stable kernels. This bug can be closed.
Thank you for testing, Bill and thank you for working on this, Larry.