On Wednesday 10 November 2010 17:04:55 Greg KH wrote:
I'm wondering why you accepted this patch despite the raised objections regarding alignment problems. [1][2]
Because at the end of that thread, it sounded like you all agreed that this patch was acceptable.
If not, then please let me know and I will revert it.
The end of the thread was that he should remove parts of the patch and resent it - or prove that all the data is 2 bytes aligned.
On Wednesday 03 November 2010 11:56:19 Sven Eckelmann wrote: [...]
I don't think they need to be two bytes aligned, but I might be wrong.
compare_ether_addr uses a two byte pointer to access 3x two bytes. This makes it necessary to have all those 3 bytes aligned to 2 byte boundaries.
[correction sent later: "6x two bytes aligned to 2 byte boundaries"]
Otherwise the compiler has to generate special instructions on architectures which don't support loads on non-aligned addresses. Usually he doesn't do it unless he has some indications that it is necessary (__attribute__ ((packed)) for example).
There is also documentation available on that topic in Documentation/unaligned-memory-access.txt
And maybe it is good to use is_broadcast_ether_addr, but leave compare_ether_addr part open (or prove that we always have those two operands correctly aligned).
[...]
None of that happened yet. And yes, this wasn't the last mail we sent to him to get some response - no answer till now.
Best regards, Sven