[PATCH v4 0/4] SPL NVMe support

[mchitale at ventanamicro.com]

On Mon, 2023-07-17 at 11:12 -0400, Tom Rini wrote:
> On Mon, Jul 17, 2023 at 01:39:52PM +0530, mchitale at ventanamicro.com
> wrote:
> > On Wed, 2023-07-12 at 13:12 -0400, Tom Rini wrote:
> > > On Wed, Jul 12, 2023 at 03:27:45PM +0200, Heinrich Schuchardt
> > > wrote:
> > > > On 12.07.23 15:06, mchitale at ventanamicro.com wrote:
> > > > > Hi Tom,
> > > > > 
> > > > > On Tue, 2023-06-20 at 09:37 -0400, Tom Rini wrote:
> > > > > > On Sat, 03 Jun 2023 19:32:52 +0530, Mayuresh Chitale wrote:
> > > > > > 
> > > > > > > This patchset adds support to load images of the SPL's
> > > > > > > next
> > > > > > > booting
> > > > > > > stage from a NVMe device.
> > > > > > > 
> > > > > > > Changes in v4:
> > > > > > > - Drop patch 4
> > > > > > > - Modify patch 2 to use generic fs.h APIs
> > > > > > > 
> > > > > > > [...]
> > > > > > 
> > > > > > With one change, which is that the "disk/part.c" in 4/4
> > > > > > were
> > > > > > not
> > > > > > required for
> > > > > > any platform in tree and also broke testcases, and so was
> > > > > > dropped,
> > > > > > this has now
> > > > > > been applied to u-boot/next. If you can explain a bit more
> > > > > > what
> > > > > > the
> > > > > > problem you
> > > > > > had was, we can look in to it. I suspect you need to test
> > > > > > for
> > > > > > not
> > > > > > SPL_ENV_SUPPORT  but ENV_SUPPORT itself.
> > > > > > 
> > > > > Thanks.
> > > > > When SPL_NVME is enabled the build breaks with the following
> > > > > error:
> > > > > riscv64-unknown-linux-gnu-ld.bfd: disk/part.o: in function
> > > > > `blk_get_device_part_str':
> > > > > u-boot/disk/part.c:473: undefined reference to `env_get'
> > > > > make[2]: *** [u-boot/scripts/Makefile.spl:527: spl/u-boot-
> > > > > spl]
> > > > > Error 1
> > > > > make[1]: *** [/u-boot/Makefile:2053: spl/u-boot-spl] Error 2
> > > > > 
> > > > > One possible fix is:
> > > > > 
> > > > > if ((!IS_ENABLED(CONFIG_SPL) &&
> > > > > IS_ENABLED(CONFIG_ENV_SUPPORT))
> > > > >         (IS_ENABLED(CONFIG_SPL) &&
> > > > > IS_ENABLED(CONFIG_SPL_ENV_SUPPORT)))
> > > > > 	if (!dev_part_str || !strlen(dev_part_str)
> > > > > > > !strcmp(dev_part_str, "-"))
> > > > > 	dev_part_str = env_get("bootdevice");
> > > > > 
> > > > > 
> > > > 
> > > > I think CONFIG_SPL_ENV_SUPPORT should depend on
> > > > CONFIG_ENV_SUPPORT
> > > > in
> > > > common/spl/Kconfig.
> > > 
> > > Not strictly, but checking for CONFIG_IS_ENABLED(ENV_SUPPORT)
> > > should
> > > do
> > > what's desired here?
> > 
> > When SPL_NVME & SPL_BLK_FS is enabled, the spl_blk_fs driver calls
> > fs_set_blk_dev to the set the device & partition before accessing
> > it
> > and fs_set_blk_dev internally tries to get the device & partition
> > from
> > the bootdevice env variable if it was not passed by the caller.
> > However
> > for SPL build, when SPL_ENV_SUPPORT is not enabled nothing provides
> > env_get and hence the build fails.
> 
> OK.  So in the code we should be able to test with
> CONFIG_IS_ENABLED(ENV_SUPPORT) is that will be true for ((SPL and
> SPL environment support) or (Main U-Boot and environment support) or
> (TPL and TPL env)).

Thanks! I will send the patch with this change.
>