On Thu, May 16, 2019 at 02:14:19PM -0500, Andreas Dannenberg wrote: > Hi Tom, > > On Thu, May 16, 2019 at 10:47:30AM -0500, Andreas Dannenberg wrote: > > Hi Tom, > > > > On Wed, May 15, 2019 at 05:50:31PM -0400, Tom Rini wrote: > > > On Wed, May 15, 2019 at 04:39:22PM -0500, Andreas Dannenberg wrote: > > > > Hi Tom, > > > > > > > > On Wed, May 15, 2019 at 11:17:22AM -0400, Tom Rini wrote: > > > > > On Tue, May 07, 2019 at 12:25:33PM -0500, Andreas Dannenberg wrote: > > > > > > > > > > > Introduce a framework that allows loading the System Firmware > > > > > > (SYSFW) > > > > > > binary as well as the associated configuration data from an image > > > > > > tree > > > > > > blob named "sysfw.itb" from an FS-based MMC boot media or from an > > > > > > MMC > > > > > > RAW mode partition or sector. > > > > > > > > > > > > To simplify the handling of and loading from the different boot > > > > > > media > > > > > > we tap into the existing U-Boot SPL framework usually used for > > > > > > loading > > > > > > U-Boot by building on an earlier commit that exposes some of that > > > > > > functionality. > > > > > > > > > > > > Note that this initial implementation only supports FS and RAW-based > > > > > > eMMC/SD card boot. > > > > > [snip] > > > > > > diff --git a/arch/arm/mach-k3/Kconfig b/arch/arm/mach-k3/Kconfig > > > > > > index e677a2e01b..f1731dda58 100644 > > > > > > --- a/arch/arm/mach-k3/Kconfig > > > > > > +++ b/arch/arm/mach-k3/Kconfig > > > > > > @@ -58,6 +58,46 @@ config SYS_K3_BOOT_CORE_ID > > > > > > int > > > > > > default 16 > > > > > > > > > > > > +config K3_LOAD_SYSFW > > > > > > + bool > > > > > > + depends on SPL > > > > > > + default n > > > > > > > > > > 'n' is already default, you can drop this. > > > > > > > > Ok sure. > > > > > > > > > > > > > > [snip] > > > > > > +config K3_SYSFW_IMAGE_SIZE_MAX > > > > > > + int "Amount of memory dynamically allocated for loading SYSFW > > > > > > blob" > > > > > > + depends on K3_LOAD_SYSFW > > > > > > + default 269000 > > > > > > + help > > > > > > + Amount of memory reserved through dynamic allocation at > > > > > > runtime for > > > > > > + loading the combined System Firmware and configuration image > > > > > > tree > > > > > > + blob. Keep it as tight as possible, as this directly affects > > > > > > the > > > > > > + overall SPL memory footprint. > > > > > > > > > > This is missing a unit, and is 'int' really the best choice here (and > > > > > really, I guess, 262.6KiB as a default) ? > > > > I think 'int' is a more natural fit as this is to reserve memory for the > > 'sysfw.itb' blob we load from media. This blob is build outside U-Boot > > using the "System Firmware Image Generator" project [1], and if after > > build somebody does an 'll' in the build directory the file size would > > show up as decimal, and that's what needs to fit inside > > K3_SYSFW_IMAGE_SIZE_MAX. > > > > > > Will add a unit. > > > > > > > > As for the specific value, in our R5 SPL memory is very very tight. The > > > > value used here is basically used for a malloc(), and any extra byte > > > > allocated will be "wasted" and not available for stack etc. Also our > > > > SYSFW image that is loaded is fixed size (except some +/-100 odd bytes > > > > if the configuration data is changed that's part of the same SYSFW.ITB > > > > blob), so a rather tailored size makes sense here. > > > > > > Right, that makes sense. But how did you get to that > > > not-exactly-"round" number rather than say 0x41A00 or some other > > > slightly smaller value in hex? If 0x41AC8 is right, then, OK, it's > > > right. It just seems strange at first. > > > > The sysfw.itb blob consumed by the SYSFW loader comprises a fixed > > component (the SYSFW itself), plus a few smaller chunks of variable data > > containing user configuration data. During initial development the > > combined size of this ITB was about 264,900 bytes, and 4KB was added as > > slack to accommodate configuration changes, resulting in what looks like > > the arbitrary number under discussion. The sysfw.itb blob has actually > > since grown in size but there is still plenty of space for it to further > > grow if needed. Since we use this value already in production since last > > year with no issues I'd rather leave it alone (due to using stack based > > malloc in R5 SPL increasing that value would directly decrease available > > stack space with our R5 SPL memory layout where the stack is growing > > down towards the R5 SPL image itself). > > Small clarification, the size of the early malloc pool is of course > controlled by CONFIG_SYS_MALLOC_F_LEN with K3_SYSFW_IMAGE_SIZE_MAX using > the majority of it. However CONFIG_SYS_MALLOC_F_LEN is dimensioned such > to "tightly wrap" both K3_SYSFW_IMAGE_SIZE_MAX and other users of the > early malloc pool (fat fs is a major one), so the principle is the same. > I'd rather leave K3_SYSFW_IMAGE_SIZE_MAX as it stands. > > Some more details on the memory usage to follow...
That's a very helpful explanation, thanks! -- Tom
signature.asc
Description: PGP signature
_______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot