Re: [PATCH v5 8/8] remoteproc: Adapt coredump to generate correct elf type

[ClÃment Leger]

Hi Bjorn,

----- On 10 Mar, 2020, at 00:57, Bjorn Andersson bjorn.andersson@xxxxxxxxxx wrote:

> On Mon 09 Mar 13:32 PDT 2020, Mathieu Poirier wrote:
> 
>> On Mon, Mar 02, 2020 at 10:39:02AM +0100, Clement Leger wrote:
>> > Now that remoteproc can load an elf64, coredump elf class should be
>> > the same as the loaded elf class. In order to do that, add a
>> > elf_class field to rproc with default values. If an elf is loaded
>> > successfully, this field will be updated with the loaded elf class.
>> > Then, the coredump core code has been modified to use the generic elf
>> > macro in order to create an elf file with correct class.
>> > 
>> > Signed-off-by: Clement Leger <cleger@xxxxxxxxx>
>> > ---
>> >  drivers/remoteproc/remoteproc_core.c       | 67 ++++++++++++++++--------------
>> >  drivers/remoteproc/remoteproc_elf_loader.c |  3 ++
>> >  include/linux/remoteproc.h                 |  1 +
>> >  3 files changed, 39 insertions(+), 32 deletions(-)
>> > 
>> > diff --git a/drivers/remoteproc/remoteproc_core.c
>> > b/drivers/remoteproc/remoteproc_core.c
>> > index b932a64a2be2..f923355aa3f9 100644
>> > --- a/drivers/remoteproc/remoteproc_core.c
>> > +++ b/drivers/remoteproc/remoteproc_core.c
>> > @@ -38,6 +38,7 @@
>> >  #include <linux/platform_device.h>
>> >  
>> >  #include "remoteproc_internal.h"
>> > +#include "remoteproc_elf_helpers.h"
>> >  
>> >  #define HIGH_BITS_MASK 0xFFFFFFFF00000000ULL
>> >  
>> > @@ -1566,20 +1567,21 @@ EXPORT_SYMBOL(rproc_coredump_add_custom_segment);
>> >  static void rproc_coredump(struct rproc *rproc)
>> >  {
>> >  	struct rproc_dump_segment *segment;
>> > -	struct elf32_phdr *phdr;
>> > -	struct elf32_hdr *ehdr;
>> > +	void *phdr;
>> > +	void *ehdr;
>> >  	size_t data_size;
>> >  	size_t offset;
>> >  	void *data;
>> >  	void *ptr;
>> > +	u8 class = rproc->elf_class;
>> >  	int phnum = 0;
>> >  
>> >  	if (list_empty(&rproc->dump_segments))
>> >  		return;
>> >  
>> > -	data_size = sizeof(*ehdr);
>> > +	data_size = elf_size_of_hdr(class);
>> >  	list_for_each_entry(segment, &rproc->dump_segments, node) {
>> > -		data_size += sizeof(*phdr) + segment->size;
>> > +		data_size += elf_size_of_phdr(class) + segment->size;
>> >  
>> >  		phnum++;
>> >  	}
>> > @@ -1590,33 +1592,33 @@ static void rproc_coredump(struct rproc *rproc)
>> >  
>> >  	ehdr = data;
>> >  
>> > -	memset(ehdr, 0, sizeof(*ehdr));
>> > -	memcpy(ehdr->e_ident, ELFMAG, SELFMAG);
>> > -	ehdr->e_ident[EI_CLASS] = ELFCLASS32;
>> > -	ehdr->e_ident[EI_DATA] = ELFDATA2LSB;
>> > -	ehdr->e_ident[EI_VERSION] = EV_CURRENT;
>> > -	ehdr->e_ident[EI_OSABI] = ELFOSABI_NONE;
>> > -	ehdr->e_type = ET_CORE;
>> > -	ehdr->e_machine = EM_NONE;
>> > -	ehdr->e_version = EV_CURRENT;
>> > -	ehdr->e_entry = rproc->bootaddr;
>> > -	ehdr->e_phoff = sizeof(*ehdr);
>> > -	ehdr->e_ehsize = sizeof(*ehdr);
>> > -	ehdr->e_phentsize = sizeof(*phdr);
>> > -	ehdr->e_phnum = phnum;
>> > -
>> > -	phdr = data + ehdr->e_phoff;
>> > -	offset = ehdr->e_phoff + sizeof(*phdr) * ehdr->e_phnum;
>> > +	memset(ehdr, 0, elf_size_of_hdr(class));
>> > +	/* e_ident field is common for both elf32 and elf64 */
>> > +	elf_hdr_init_ident(ehdr, class);
>> > +
>> > +	elf_hdr_set_e_type(class, ehdr, ET_CORE);
>> > +	elf_hdr_set_e_machine(class, ehdr, EM_NONE);
>> > +	elf_hdr_set_e_version(class, ehdr, EV_CURRENT);
>> > +	elf_hdr_set_e_entry(class, ehdr, rproc->bootaddr);
>> > +	elf_hdr_set_e_phoff(class, ehdr, elf_size_of_hdr(class));
>> > +	elf_hdr_set_e_ehsize(class, ehdr, elf_size_of_hdr(class));
>> > +	elf_hdr_set_e_phentsize(class, ehdr, elf_size_of_phdr(class));
>> > +	elf_hdr_set_e_phnum(class, ehdr, phnum);
>> > +
>> > +	phdr = data + elf_hdr_get_e_phoff(class, ehdr);
>> > +	offset = elf_hdr_get_e_phoff(class, ehdr);
>> > +	offset += elf_size_of_phdr(class) * elf_hdr_get_e_phnum(class, ehdr);
>> > +
>> >  	list_for_each_entry(segment, &rproc->dump_segments, node) {
>> > -		memset(phdr, 0, sizeof(*phdr));
>> > -		phdr->p_type = PT_LOAD;
>> > -		phdr->p_offset = offset;
>> > -		phdr->p_vaddr = segment->da;
>> > -		phdr->p_paddr = segment->da;
>> > -		phdr->p_filesz = segment->size;
>> > -		phdr->p_memsz = segment->size;
>> > -		phdr->p_flags = PF_R | PF_W | PF_X;
>> > -		phdr->p_align = 0;
>> > +		memset(phdr, 0, elf_size_of_phdr(class));
>> > +		elf_phdr_set_p_type(class, phdr, PT_LOAD);
>> > +		elf_phdr_set_p_offset(class, phdr, offset);
>> > +		elf_phdr_set_p_vaddr(class, phdr, segment->da);
>> > +		elf_phdr_set_p_paddr(class, phdr, segment->da);
>> > +		elf_phdr_set_p_filesz(class, phdr, segment->size);
>> > +		elf_phdr_set_p_memsz(class, phdr, segment->size);
>> > +		elf_phdr_set_p_flags(class, phdr, PF_R | PF_W | PF_X);
>> > +		elf_phdr_set_p_align(class, phdr, 0);
>> >  
>> >  		if (segment->dump) {
>> >  			segment->dump(rproc, segment, data + offset);
>> > @@ -1632,8 +1634,8 @@ static void rproc_coredump(struct rproc *rproc)
>> >  			}
>> >  		}
>> >  
>> > -		offset += phdr->p_filesz;
>> > -		phdr++;
>> > +		offset += elf_phdr_get_p_filesz(class, phdr);
>> > +		phdr += elf_size_of_phdr(class);
>> >  	}
>> >  
>> >  	dev_coredumpv(&rproc->dev, data, data_size, GFP_KERNEL);
>> > @@ -2031,6 +2033,7 @@ struct rproc *rproc_alloc(struct device *dev, const char
>> > *name,
>> >  	rproc->name = name;
>> >  	rproc->priv = &rproc[1];
>> >  	rproc->auto_boot = true;
>> > +	rproc->elf_class = ELFCLASS32;
>> 
>> I would initialise this to ELFCLASSNONE to make sure that if a platform driver
>> overwrites rproc_elf_load_segments or doesn't provide one, we don't falsely
>> deduce the elf type.  It goes without saying that if elf_class == ELFCLASSNONE,
>> a coredump is not generated.
>> 
> 
> I like the idea of making the choice explicit, perhaps even more
> explicit than the assumption that the coredumps should be of the same
> type as the ELF loaded. Note that it's different consumers of the two
> ELF files.
> 
>> Unless you think this is a seriously bad idea or Bjorn over rules me,

Ok, I tried to do the equivalent of what was existing (ie elf32 by default).
But IMHO, letting the driver choose the elf type is a better idea.

>> 
>> Reviewed-by: Mathieu Poirier <mathieu.poirier@xxxxxxxxxx>
>> 
> 
> Not sure if it count as "over ruling", I accept your suggestion but used
> your R-b to merge the patch as is, no need to hold this up any longer.
> 
> Clement, can you please follow up with a patch implementing this (don't
> forget that the qcom drivers doesn't use rproc_elf_load_segments())

I was going to send a v7, please tell me if you want to hold it a bit more.
If not, I will address the remaining comments in next commits.

> 
> Thanks Clement and thanks for the reviews Mathieu.
> 
> Regards,
> Bjorn
> 
>> Thanks,
>> Mathieu
>> 
>> >  
>> >  	device_initialize(&rproc->dev);
>> >  	rproc->dev.parent = dev;
>> > diff --git a/drivers/remoteproc/remoteproc_elf_loader.c
>> > b/drivers/remoteproc/remoteproc_elf_loader.c
>> > index 4869fb7d8fe4..16e2c496fd45 100644
>> > --- a/drivers/remoteproc/remoteproc_elf_loader.c
>> > +++ b/drivers/remoteproc/remoteproc_elf_loader.c
>> > @@ -248,6 +248,9 @@ int rproc_elf_load_segments(struct rproc *rproc, const
>> > struct firmware *fw)
>> >  			memset(ptr + filesz, 0, memsz - filesz);
>> >  	}
>> >  
>> > +	if (ret == 0)
>> > +		rproc->elf_class = class;
>> > +
>> >  	return ret;
>> >  }
>> >  EXPORT_SYMBOL(rproc_elf_load_segments);
>> > diff --git a/include/linux/remoteproc.h b/include/linux/remoteproc.h
>> > index 1683d6c386a6..ed127b2d35ca 100644
>> > --- a/include/linux/remoteproc.h
>> > +++ b/include/linux/remoteproc.h
>> > @@ -514,6 +514,7 @@ struct rproc {
>> >  	bool auto_boot;
>> >  	struct list_head dump_segments;
>> >  	int nb_vdev;
>> > +	u8 elf_class;
>> >  };
>> >  
>> >  /**
>> > --
>> > 2.15.0.276.g89ea799