[PATCH v9 07/11] arm64: kexec_file: add crash dump support

[takahiro.akashi@linaro.org (AKASHI Takahiro)]

Hi Rob,

On Fri, May 18, 2018 at 10:35:52AM -0500, Rob Herring wrote:
> On Tue, May 15, 2018 at 06:12:59PM +0100, James Morse wrote:
> > Hi guys,
> > 
> > (CC: +RobH, devicetree list)
> 
> Thanks.
> 
> > On 25/04/18 07:26, AKASHI Takahiro wrote:
> > > Enabling crash dump (kdump) includes
> > > * prepare contents of ELF header of a core dump file, /proc/vmcore,
> > >   using crash_prepare_elf64_headers(), and
> > > * add two device tree properties, "linux,usable-memory-range" and
> > >   "linux,elfcorehdr", which represent repsectively a memory range
> > >   to be used by crash dump kernel and the header's location
> 
> BTW, I intend to move existing parsing these out of the arch code. 
> Please don't add more DT handling to arch/ unless it is *really* arch 
> specific. I'd assume that the next arch to add kexec support will use 
> these bindings instead of the powerpc way.

So do you expect all the fdt-related stuff in my current implementation
for arm64 to be put into libfdt, or at least drivers/of, from the beginning?

I'm not sure how arch-specific the properties here are. For instance,
it is only arm64 that uses "linux,usable-memory-range" right now but
if some other arch follows, it is no more arch-specific.
# I remember that you didn't like this property :)

> > kexec_file_load() on arm64 needs to be able to create a prop encoded array to
> > the FDT, but there doesn't appear to be a libfdt helper to do this.
> > 
> > Akashi's code below adds fdt_setprop_range() to the arch code, and duplicates
> > bits of libfdt_internal.h to do the work.
> > 
> > How should this be done? I'm assuming this is something we need a new API in
> > libfdt.h for. How do these come about, and is there an interim step we can use
> > until then?
> 
> Submit patches to upstream dtc and then we can pull it in. Ahead of that 
> you can add it to drivers/of/fdt.c (or maybe fdt_address.c because 
> that's really what this is dealing with).

OK, I'm going to try to follow your suggestion.

> libfdt has only recently gained the beginnings of address handling.
> 
> > 
> > Thanks!
> > 
> > James
> > 
> > > diff --git a/arch/arm64/kernel/machine_kexec_file.c b/arch/arm64/kernel/machine_kexec_file.c
> > > index 37c0a9dc2e47..ec674f4d267c 100644
> > > --- a/arch/arm64/kernel/machine_kexec_file.c
> > > +++ b/arch/arm64/kernel/machine_kexec_file.c
> > > @@ -76,6 +81,78 @@ int arch_kexec_walk_mem(struct kexec_buf *kbuf,
> > >  	return ret;
> > >  }
> > >  
> > > +static int __init arch_kexec_file_init(void)
> > > +{
> > > +	/* Those values are used later on loading the kernel */
> > > +	__dt_root_addr_cells = dt_root_addr_cells;
> > > +	__dt_root_size_cells = dt_root_size_cells;
> 
> I intend to make dt_root_*_cells private, so don't add another user 
> outside of drivers/of/.

Once cells_size_fitted() moves to drivers/of, there will be no users.

> > > +
> > > +	return 0;
> > > +}
> > > +late_initcall(arch_kexec_file_init);
> > > +
> > > +#define FDT_ALIGN(x, a)	(((x) + (a) - 1) & ~((a) - 1))
> > > +#define FDT_TAGALIGN(x)	(FDT_ALIGN((x), FDT_TAGSIZE))
> > > +
> > > +static int fdt_prop_len(const char *prop_name, int len)
> > > +{
> > > +	return (strlen(prop_name) + 1) +
> > > +		sizeof(struct fdt_property) +
> > > +		FDT_TAGALIGN(len);
> > > +}
> > > +
> > > +static bool cells_size_fitted(unsigned long base, unsigned long size)
> 
> I can't imagine this would happen. However, when this is moved to 
> drivers/of/ or dtc, these need to be u64 types to work on 32-bit.

OK.

> > > +	/* if *_cells >= 2, cells can hold 64-bit values anyway */
> > > +	if ((__dt_root_addr_cells == 1) && (base >= (1ULL << 32)))
> > > +		return false;
> > > +
> > > +	if ((__dt_root_size_cells == 1) && (size >= (1ULL << 32)))
> > > +		return false;
> > > +
> > > +	return true;
> > > +}
> > > +
> > > +static void fill_property(void *buf, u64 val64, int cells)
> > > +{
> > > +	u32 val32;
> 
> This should be a __be32 or fdt32 type. So should buf.

OK for val32, but buf is a local pointer address.

> > > +
> > > +	if (cells == 1) {
> > > +		val32 = cpu_to_fdt32((u32)val64);
> > > +		memcpy(buf, &val32, sizeof(val32));
> > > +	} else {
> > > +		memset(buf, 0, cells * sizeof(u32) - sizeof(u64));
> > > +		buf += cells * sizeof(u32) - sizeof(u64);
> > > +
> > > +		val64 = cpu_to_fdt64(val64);
> > > +		memcpy(buf, &val64, sizeof(val64));
> 
> Look how of_read_number() is implemented. You should be able to do 
> something similar here looping and avoiding the if/else.

Ah, excellent!

> > > +	}
> > > +}
> > > +
> > > +static int fdt_setprop_range(void *fdt, int nodeoffset, const char *name,
> > > +				unsigned long addr, unsigned long size)
> 
> A very generic sounding function, but really only works on addresses in 
> children of the root node.
> 
> > > +{
> > > +	void *buf, *prop;
> > > +	size_t buf_size;
> > > +	int result;
> > > +
> > > +	buf_size = (__dt_root_addr_cells + __dt_root_size_cells) * sizeof(u32);
> > > +	prop = buf = vmalloc(buf_size);
> 
> This can go on the stack instead (and would be required to to work in 
> libfdt).

Well, I can't agree with you here since we are now in effort, as far as
I correctly understand, of purging all the variable-sized arrays on a local
stack out of the kernel code.

Thank you for your review.
-Takahiro AKASHI

> > > +	if (!buf)
> > > +		return -ENOMEM;
> > > +
> > > +	fill_property(prop, addr, __dt_root_addr_cells);
> > > +	prop += __dt_root_addr_cells * sizeof(u32);
> > > +
> > > +	fill_property(prop, size, __dt_root_size_cells);
> > > +
> > > +	result = fdt_setprop(fdt, nodeoffset, name, buf, buf_size);
> > > +
> > > +	vfree(buf);
> > > +
> > > +	return result;
> > > +}
> > > +
> > >  static int setup_dtb(struct kimage *image,
> > >  		unsigned long initrd_load_addr, unsigned long initrd_len,
> > >  		char *cmdline, unsigned long cmdline_len,
> > > @@ -88,10 +165,26 @@ static int setup_dtb(struct kimage *image,
> > >  	int range_len;
> > >  	int ret;
> > >  
> > > +	/* check ranges against root's #address-cells and #size-cells */
> > > +	if (image->type == KEXEC_TYPE_CRASH &&
> > > +		(!cells_size_fitted(image->arch.elf_load_addr,
> > > +				image->arch.elf_headers_sz) ||
> > > +		 !cells_size_fitted(crashk_res.start,
> > > +				crashk_res.end - crashk_res.start + 1))) {
> > > +		pr_err("Crash memory region doesn't fit into DT's root cell sizes.\n");
> > > +		ret = -EINVAL;
> > > +		goto out_err;
> > > +	}
> > > +
> > >  	/* duplicate dt blob */
> > >  	buf_size = fdt_totalsize(initial_boot_params);
> > >  	range_len = (__dt_root_addr_cells + __dt_root_size_cells) * sizeof(u32);
> > >  
> > > +	if (image->type == KEXEC_TYPE_CRASH)
> > > +		buf_size += fdt_prop_len("linux,elfcorehdr", range_len)
> > > +				+ fdt_prop_len("linux,usable-memory-range",
> > > +								range_len);
> > > +
> > >  	if (initrd_load_addr)
> > >  		buf_size += fdt_prop_len("linux,initrd-start", sizeof(u64))
> > >  				+ fdt_prop_len("linux,initrd-end", sizeof(u64));
> > > @@ -113,6 +206,23 @@ static int setup_dtb(struct kimage *image,
> > >  	if (nodeoffset < 0)
> > >  		goto out_err;
> > >  
> > > +	if (image->type == KEXEC_TYPE_CRASH) {
> > > +		/* add linux,elfcorehdr */
> > > +		ret = fdt_setprop_range(buf, nodeoffset, "linux,elfcorehdr",
> > > +				image->arch.elf_load_addr,
> > > +				image->arch.elf_headers_sz);
> > > +		if (ret)
> > > +			goto out_err;
> > > +
> > > +		/* add linux,usable-memory-range */
> > > +		ret = fdt_setprop_range(buf, nodeoffset,
> > > +				"linux,usable-memory-range",
> > > +				crashk_res.start,
> > > +				crashk_res.end - crashk_res.start + 1);
> > > +		if (ret)
> > > +			goto out_err;
> > > +	}
> > > +
> > >  	/* add bootargs */
> > >  	if (cmdline) {
> > >  		ret = fdt_setprop(buf, nodeoffset, "bootargs",
> >