Hi Wolfgang,
On Fri, 15 Mar 2019 at 02:34, Wolfgang Denk wd@denx.de wrote:
Dear Simon,
In message CAPnjgZ14Sq=OWRbM6hpqO2CLP=1eg=ftQw_1Bqu0DNe+6OxTkQ@mail.gmail.com you wrote:
I think it is a reasonable idea to allow the gd region to pass from TPL -> SPL -> U-Boot. But we'll need to remove use of CONFIG_IS_ENABLED(), or put shared things at the beginning of the structure.
Indeed. And/or split things up in "common" stuff and optional / config dependent things.
We need the concept of 'am I the first thing to run'. This is implemented in bloblist as u_boot_first_phase() - see spl.h. If this is true, we must set up the data structure. If false we must find one set up by a previous phase and use it. Bloblist handles this, but perhaps gd could as well?
I wonder why we need 4 different ways of doing basically the same thing.
First, we have GD, which exists since the dawn of U-Boot, which was intended to pass data between boot stages (by then, before and after relocation), but apparently it has never been used for passing information between SPL and U-Boot proper.
That's right. It is always zeroed these days at the start of each phase. So I think work is needed if we want to use this.
Also I don't think we can assume that gd stays in the same place through all phases of U-Boot. Probably we can keep it in the same place in TPL, SPL and U-Boot pre-relocation, then move it to SDRAM during relocation.
Then you added the bloblist thingy. It's not really clear what it's intentions are - I see the commits, but I can't find what you want to use it for or what design you have in mind. It's too complicated for passing just a few data, but apparently you find it necessary to make it secure enough that you add version, magic and checksum (which makes it necessary to have CRC32 in SPL...). Also, I wonder how the search mechanism effects boot time...
This is designed for things that need to past structs between phases. For example, verified boot may start processing in TPL and then pass information to SPL and then to U-Boot proper. This may involve quite a bit of info, so it is in a C structure. It isn't really suitable to put this entire structure in gd IMO, since that struct is pretty small.
See doc/README.bloblist:
A bloblist provides a way to store collections of binary information (blobs) in a central structure. Each record of information is assigned a tag so that its owner can find it and update it. Each record is generally described by a C structure defined by the code that owns it.
Maybe this should have more detail?
The version, magic and checksum are to ensure that the data is not corrupted by mistake, which as you know can happen very easily with fixed-position data structures. The search is pretty quick once the checksum is done, just running through a few pointers. I suppose we could make the checksum optional.
An then there is commit b0edea3c27 with the spl_handoff thing. I can't decide whether this is intended as a general feature or a separate, SPL specific mechanism. And more questions - if we pass the handoff pointer in GD, why all the effort - why don't we just make sure GD is passed properly? The fact that there is no use case at all in mainline U-Boot makes it really hard to understand your intentions.
That actually uses the bloblist, putting an SPL struct in there, intended to hold things like the SDRAM config or boot options discovered in SPL. At present we have a few ad-hoc ways of
And finally there is bootstage with it's own mechanism of information passing.
Yes, and this is ad-hoc too. I would like to move it to bloblist.
Can we not unify these, and use one common method, please?
I think we might end up with gd (once this work is done) and bloblist. For now I feel that bloblist has a purpose but let's see how it goes with the gd work.
Also consider the scenario where there is a read-only TPL programmed in manufacture that never changes, and a read-write SPL + U-Boot that can be upgraded in the field. In this case they may eventually end up being built with different versions of U-Boot. The bloblist structure is intended to handle this by at least checking that the size matches.
You also have the version field there, right? Who not (also) checking this?
Yes it should check this, at least once we actually have a version 1 :-)
Related, I feel that we should figure out how to use registers to pass addresses from SPL to U-Boot. On ARM we could use r0 to pass the value of gd, perhaps.
There is no need to. GD already has a well-defined register which has been reserved exclusively for this use - on ARM, it's R9.
Yes, but I'm not sure this is portable. E.g. on x86 this can't be used in 64-bit mode, so passing info from 32-bit SPL to 64-bit U-Boot is not possible (I believe). Using a normal register might be better, but in any case this is going to be arch-specific.
In any case, I was not even aware that gd was intended to pass info from SPL to U-Boot. It is possible that in my refactoring of board_init() some years ago I actually broken this. Once we get it running again we should add some tests and docs.
Regards, Simon