Hi Tom, On Tue, 15 Oct 2024 at 08:37, Tom Rini <[email protected]> wrote: > > On Tue, Oct 15, 2024 at 07:33:00AM -0600, Simon Glass wrote: > > > The board_init_f() function for am65x is a bit confusing, since it uses > > the variable name 'pool_addr' to hold a pointer. It then casts it to an > > address to pass to mem_alloc_init() > > > > Rename the variable and use mapmem to convert to an address. > > > > Signed-off-by: Simon Glass <[email protected]> > > --- > > > > (no changes since v1) > > > > arch/arm/mach-k3/am65x/am654_init.c | 11 ++++++----- > > 1 file changed, 6 insertions(+), 5 deletions(-) > > > > diff --git a/arch/arm/mach-k3/am65x/am654_init.c > > b/arch/arm/mach-k3/am65x/am654_init.c > > index a4f038029d7..ddcc1355210 100644 > > --- a/arch/arm/mach-k3/am65x/am654_init.c > > +++ b/arch/arm/mach-k3/am65x/am654_init.c > > @@ -8,6 +8,7 @@ > > > > #include <fdt_support.h> > > #include <init.h> > > +#include <mapmem.h> > > #include <asm/global_data.h> > > #include <asm/io.h> > > #include <spl.h> > > @@ -165,7 +166,7 @@ void board_init_f(ulong dummy) > > #if defined(CONFIG_K3_LOAD_SYSFW) || defined(CONFIG_K3_AM654_DDRSS) > > struct udevice *dev; > > size_t pool_size; > > - void *pool_addr; > > + void *pool; > > int ret; > > #endif > > /* > > @@ -204,14 +205,14 @@ void board_init_f(ulong dummy) > > * malloc pool of which we use all that's left. > > */ > > pool_size = CONFIG_VAL(SYS_MALLOC_F_LEN) - gd->malloc_ptr; > > - pool_addr = malloc(pool_size); > > - if (!pool_addr) > > + pool = malloc(pool_size); > > + if (!pool) > > panic("ERROR: Can't allocate full malloc pool!\n"); > > > > - mem_malloc_init((ulong)pool_addr, (ulong)pool_size); > > + mem_malloc_init(map_to_sysmem(pool), (ulong)pool_size); > > gd->flags |= GD_FLG_FULL_MALLOC_INIT; > > debug("%s: initialized an early full malloc pool at 0x%08lx of 0x%lx > > bytes\n", > > - __func__, (unsigned long)pool_addr, (unsigned long)pool_size); > > + __func__, (unsigned long)pool, (unsigned long)pool_size); > > /* > > * Process pinctrl for the serial0 a.k.a. WKUP_UART0 module and > > continue > > * regardless of the result of pinctrl. Do this without probing the > > This covers half of what we talked about in v2. But map_to_sysmem() is a > sandbox thing that should not be called outside of (a) arch/sandbox or > (b) common code that runs on sandbox. And honestly, we should work to > hide that abstraction as much as we can, not start introducing it more.
Yes that's right. Perhaps we should document that. We can drop this patch. Regards, Simon
