User Details
- User Since
- Apr 24 2016, 4:47 PM (552 w, 2 h)
Today
I'm not against changing default depth (32bpp seems better to have by default) but it raises two questions :
- Why does the firmware sets the default to 16bpp ?
- We still panic with the default firmware configuration ?
Fri, Nov 20
That's just hidding the real problem IMHO
Thanks for the review.
Looking at the RK3568 TRM (section 20.6.3, TDES3 read format): bits [14:0] are FL/TPL,
the length of the packet when TSE is not set,
but "when DWRR/WFQ algorithm is NOT enabled,
value written into this field is not used when TSE = 0".
The driver uses a single TX queue and does not select a scheduling algorithm,
so the hardware never looks at the value and this change has no effect.
Linux and U-Boot do write it, but there is nothing to gain here, so I am dropping the
commit for now and will bring it back if DWRR/WFQ or TSO gets enabled.
Rebase on latest code (Rebased on D61016)
Thu, Nov 19
Commited as 48fa466426d90bfd02191448d7b81e0da1acee8a
That's likely not the root of the problem, firmware likely created space for a 16bpp framebuffer but the driver is changing it to 24bpp (See https://cgit.freebsd.org/src/tree/sys/arm/broadcom/bcm2835/bcm2835_fbd.c#n88), that's why it crashes.
Wed, Nov 18
Tue, Nov 17
Mon, Nov 16
Sun, Nov 15
There is a lot of changes in this review and already a lot of comments, I suggest splitting this into :
So, I haven't followed RPI5 stuff for some time but what's the current behavior without this ? I'm pretty sure that people are using them.
Does that mean that we are currently using DTB loaded by u-boot and you want to switch to the DTB loaded by the firmware ? (like other PI iirc).
I'm not against the change but I'd like to understand more.
Thanks.
As soon as https://reviews.freebsd.org/D61539 is accepted this one can go in.
I really don't like having patches for u-boot, where those submitted upstream ?