On 18 December 2010 18:59, Hauke Mehrtens <ha...@hauke-m.de> wrote: > In 0008-SSB-Add-support-for-the-PMU-of-newer-Broadcoms.patch you are > checking for 47162 shouldn't it be 0x47162 or did Broadcom some new > strange thing.
No, this is correct, the BCM4716's chipid is 0xB83A, which translates to 47162 (0x47162 wouldn't fit as it's only a 16 bit long field). It seems Broadcom uses 4 digit chip ids as hex values, but 5 digit ones as is (iirc it's the same with the wifi chip ids like the 43222). > For the big patches it would be easier to review if one patch just > reorders the code, like putting it into the new functions in > 0002-SSB-Modify-core-scan.patch and an extra patch for adding some new > features or for doing some real changes. This is exactly my plan, but for I now I just took Bernhard's original patches and cleaned up the white space a bit so that git doesn't complain anymore. For submitting them eventually upstream these need to be reordered and more nicely commented. >> Also when testing you might want to remove the >> "016-MIPS-BCM47xx-Remove-CFE-console.patch" to be able to enable early >> printk. > You have to remove this patch for early printks did not > 014-MIPS-BCM47xx-Setup-and-register-serial-early.patch helped? Or does > this needed some functionality in ssb that is not implemented for these > devices? "016-MIPS-BCM47xx-Remove-CFE-console.patch" removes the SYS_HAS_EARLY_PRINTK from BCM47XX, so at least for me this made the console unavailable at least until the SSB registering stage (I didn't check again after I removed the patch and enabled early printk in the kernel config and then later got it go past it). >> @Hauke: I had to remove the revision check for the num_cores detection >> since reading the register hangs (at least) the BCM4716. >> If we keep that part of the code, we might want to first check if >> there's a known chipid => num_cores mapping, and if not read the >> number of cores from the register without checking the core revision. > Hopefully there is some other register that is indicating if we could > read out this register or not. I haven't checked, but we might be able to check whether we have an AI or SSB common core, after which we should be able to read out this register for SSB cores. > Are there big changes between 4716 and 47162? As far as I can tell the 47162 is a real BCM4716 while the 0x4716 is actually a BCM4718 (the biggest one for now until the BCM4748 gets released). Jonas _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel
