| Message ID | 1370010272-25520-2-git-send-email-lee.jones@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Ah, cack! One of the pitfalls of autonomously using the '-a' flag. I'll re-spin.
On Friday 31 May 2013 15:24:32 Lee Jones wrote: > In order to utilise all of the memory located on ux500 based devices > we have to enable HIGHMEM. Without it the kernel truncates memory down > to what's left after the PAGE_OFFSET has been applied, which doesn't > leave an awful lot, especially if we're running large rootfs' such > as full Linux desktop distributions or Android. > > Signed-off-by: Lee Jones <lee.jones@linaro.org> How much memory do those devices have? Enabling HIGHMEM is quite expensive. > diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c > index 5439925..808be06 100644 > --- a/drivers/of/fdt.c > +++ b/drivers/of/fdt.c You should probably resend the patch without the second half that got in it by accident. ARnd
On Fri, 31 May 2013, Arnd Bergmann wrote: > On Friday 31 May 2013 15:24:32 Lee Jones wrote: > > In order to utilise all of the memory located on ux500 based devices > > we have to enable HIGHMEM. Without it the kernel truncates memory down > > to what's left after the PAGE_OFFSET has been applied, which doesn't > > leave an awful lot, especially if we're running large rootfs' such > > as full Linux desktop distributions or Android. > > > > Signed-off-by: Lee Jones <lee.jones@linaro.org> > > How much memory do those devices have? Enabling HIGHMEM is quite expensive. Another 1GB. > > diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c > > index 5439925..808be06 100644 > > --- a/drivers/of/fdt.c > > +++ b/drivers/of/fdt.c > > You should probably resend the patch without the second half that got > in it by accident. I did already, thanks.
On Friday 31 May 2013 16:18:50 Lee Jones wrote: > On Fri, 31 May 2013, Arnd Bergmann wrote: > > > On Friday 31 May 2013 15:24:32 Lee Jones wrote: > > > In order to utilise all of the memory located on ux500 based devices > > > we have to enable HIGHMEM. Without it the kernel truncates memory down > > > to what's left after the PAGE_OFFSET has been applied, which doesn't > > > leave an awful lot, especially if we're running large rootfs' such > > > as full Linux desktop distributions or Android. > > > > > > Signed-off-by: Lee Jones <lee.jones@linaro.org> > > > > How much memory do those devices have? Enabling HIGHMEM is quite expensive. > > Another 1GB. You mean you have 2GB total or 1.75GB? You could just use CONFIG_VMSPLIT_2G to an extra GB compared to the default, which is what vexpress, imx and s5p already do, but you still have to put the vmalloc area in the low 2GB then. Arnd
On Fri, 31 May 2013, Arnd Bergmann wrote: > On Friday 31 May 2013 16:18:50 Lee Jones wrote: > > On Fri, 31 May 2013, Arnd Bergmann wrote: > > > > > On Friday 31 May 2013 15:24:32 Lee Jones wrote: > > > > In order to utilise all of the memory located on ux500 based devices > > > > we have to enable HIGHMEM. Without it the kernel truncates memory down > > > > to what's left after the PAGE_OFFSET has been applied, which doesn't > > > > leave an awful lot, especially if we're running large rootfs' such > > > > as full Linux desktop distributions or Android. > > > > > > > > Signed-off-by: Lee Jones <lee.jones@linaro.org> > > > > > > How much memory do those devices have? Enabling HIGHMEM is quite expensive. > > > > Another 1GB. > > You mean you have 2GB total or 1.75GB? You could just use We have 2GB in total. > CONFIG_VMSPLIT_2G to an extra GB compared to the default, which > is what vexpress, imx and s5p already do, but you still have to > put the vmalloc area in the low 2GB then. Isn't this for 4GB systems?
diff --git a/arch/arm/configs/u8500_defconfig b/arch/arm/configs/u8500_defconfig index 2d322d2..38d633a 100644 --- a/arch/arm/configs/u8500_defconfig +++ b/arch/arm/configs/u8500_defconfig @@ -1,3 +1,4 @@ +CONFIG_HIGHMEM=y # CONFIG_SWAP is not set CONFIG_SYSVIPC=y CONFIG_NO_HZ=y diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c index 5439925..808be06 100644 --- a/drivers/of/fdt.c +++ b/drivers/of/fdt.c @@ -43,7 +43,7 @@ void *of_fdt_get_property(struct boot_param_header *blob, do { u32 tag = be32_to_cpup((__be32 *)p); - u32 sz, noff, i; + u32 sz, noff; const char *nstr; p += 4; @@ -66,10 +66,6 @@ void *of_fdt_get_property(struct boot_param_header *blob, if (strcmp(name, nstr) == 0) { if (size) *size = sz; - - print_hex_dump(KERN_ERR, "node: ", DUMP_PREFIX_ADDRESS, - 16, 1, (__be32 *)node, p - node, true); - return (void *)p; } p += sz; @@ -624,68 +620,38 @@ int __init early_init_dt_scan_memory(unsigned long node, const char *uname, __be32 *reg, *endp; unsigned long l; - printk("LEE: ******************************************* \n"); - printk("LEE: ******************************************* \n"); - printk("LEE: ******************************************* \n"); - printk("LEE: ******************************************* \n"); - printk("LEE: ******************************************* \n"); - printk("LEE: ******************************************* \n"); - printk("LEE: ******************************************* \n"); - /* We are scanning "memory" nodes only */ if (type == NULL) { - printk("LEE: 1\n"); /* * The longtrail doesn't have a device_type on the * /memory node, so look for the node called /memory@0. */ - if (depth != 1 || strcmp(uname, "memory@0") != 0) { - printk("LEE: 2\n"); + if (depth != 1 || strcmp(uname, "memory@0") != 0) return 0; - } - } else if (strcmp(type, "memory") != 0) { - printk("LEE: 3\n"); + } else if (strcmp(type, "memory") != 0) return 0; - } reg = of_get_flat_dt_prop(node, "linux,usable-memory", &l); - if (reg == NULL) { + if (reg == NULL) reg = of_get_flat_dt_prop(node, "reg", &l); - printk("LEE: This is correct \n"); - } - if (reg == NULL) { - printk("LEE: 4\n"); - return 0; - } - - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); - printk("LEE: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! \n"); + if (reg == NULL) + return 0; endp = reg + (l / sizeof(__be32)); - printk("LEE: reg: 0x%p\n", reg); - - printk("memory scan node %s, reg size %ld, data: %x %x %x %x,\n", + pr_debug("memory scan node %s, reg size %ld, data: %x %x %x %x,\n", uname, l, reg[0], reg[1], reg[2], reg[3]); - printk("extras: %x %x %x %x,\n", reg[4], reg[5], reg[6], reg[7]); - while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) { u64 base, size; base = dt_mem_next_cell(dt_root_addr_cells, ®); size = dt_mem_next_cell(dt_root_size_cells, ®); - printk(" - %llx , %llx\n", (unsigned long long)base, - (unsigned long long)size); - if (size == 0) continue; + pr_debug(" - %llx , %llx\n", (unsigned long long)base, + (unsigned long long)size); early_init_dt_add_memory_arch(base, size); }
In order to utilise all of the memory located on ux500 based devices we have to enable HIGHMEM. Without it the kernel truncates memory down to what's left after the PAGE_OFFSET has been applied, which doesn't leave an awful lot, especially if we're running large rootfs' such as full Linux desktop distributions or Android. Signed-off-by: Lee Jones <lee.jones@linaro.org> --- arch/arm/configs/u8500_defconfig | 1 + drivers/of/fdt.c | 52 +++++++------------------------------- 2 files changed, 10 insertions(+), 43 deletions(-)