Michael, Let me explain a little more completely what I am thinking, and what I have done. As I see it, the 3 most important problems not yet solved by the bcm43xx code (in either branch) are (1) the inability to transmit at higher than 11 Mb/s, (2) the difficulties with at least the 4318 chips, and perhaps other chips as well, in a setting with several active APs, and (3) the lack of PCI-E support. I am not in a position to do much about (1) or (2), so I am trying to help with PCI-E. From reading the support forums and mailing lists, I know that a lot of people are reluctant to go to the effort of implementing and learning git; therefore, the d80211 tree is effectively off limits for them. As a result, I decided to find a willing tester (which I have), develop a patch for Linus's tree that would support PCI-E, and send that patch to the tester privately. When I started this project, I discovered that the only copy of Stefano's PCI-E patches that I could access had corrupted white space. All tabs had been replaced by spaces. I sent mail to him requesting a new copy, but as you know, he has been off-line. Accordingly, I checked to see if the ssb changes in d80211 could be backported. To do this, I looked at the differences between the softmac and d80211 versions of bcm43xx, figured out what was different, and modified softmac accordingly. It took a bit of effort, but that is the version that is now working. The code for ssb was copied from the latest wireless-dev.
Michael Buesch wrote: > On Friday 01 September 2006 18:00, Larry Finger wrote: > > It is not sent for inclusion into mainline and I am not sure how > to handle this. if we simply merge it into mainline, we will end > up with all sorts of horrible and hard to resolve rejects in the > wireless-dev tree, where the patch is merged. The version of ssb that I am using is exactly the one in wireless-dev.git now. There are _no_ differences in ssb.c nor ssb.h. I am using an earlier version of Kconfig than are you, but that is the only difference. Merging should be a non-issue. > But besides that, I don't really see a benefit from using ssb > with bcm43xx-softmac. Porting bcm43xx-softmac to ssb has a > potential to make perfectly stable and well tested code unstable. I grant you this point, but as stated earlier, this seemed to be my only way to get the PCI-E stuff into the code. I do not find bcm43xx-softmac to be particularly stable. It does not crash the machine, but the interface goes off-line 6-10 times per day, and I have to give an ifdown/ifup sequence to restore operation. It went off-line while I was writing this message. The ssb-bcm43xx-softmac version is no worse than standard wireless-2.6. > And other than that, I have big changes queued for ssb. These changes > _will_ horribly conflict with the bcm43xx-softmac port. And this is > really nontrivial to resolve and to get right. Here you have me. Once my tester and I discover what changes are needed to get PCI-E working with my bastard system, I'll submit the appropriate patches to you. Thanks, Larry _______________________________________________ Bcm43xx-dev mailing list Bcm43xx-dev@lists.berlios.de https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
