[PATCH 7/7] efi: write volatile EFI variables used by systemd

[Sascha Hauer]

On Fri, Jul 17, 2015 at 09:05:40AM +0200, Michael Olbrich wrote:
> On Thu, Jul 16, 2015 at 10:08:48PM +0200, Sascha Hauer wrote:
> > On Thu, Jul 16, 2015 at 10:43:56AM +0200, Michael Olbrich wrote:
> > > LoaderTimeInitUSec and LoaderTimeExecUSec are used e.g. in systemd-analyze
> > > to calculate the time spent in the firmare and barebox.
> > > 
> > > LoaderDevicePartUUID is used to mount the EFI partition to /boot.
> > > 
> > > Signed-off-by: Michael Olbrich <m.olbrich at pengutronix.de>
> > > ---
> > >  arch/efi/efi/efi-image.c |  5 +++++
> > >  arch/efi/efi/efi.c       | 14 ++++++++++++++
> > >  common/efi-guid.c        |  1 +
> > >  include/efi.h            |  5 +++++
> > >  4 files changed, 25 insertions(+)
> > > 
> > > diff --git a/arch/efi/efi/efi-image.c b/arch/efi/efi/efi-image.c
> > > index f41322744bbc..89b712baab1f 100644
> > > --- a/arch/efi/efi/efi-image.c
> > > +++ b/arch/efi/efi/efi-image.c
> > > @@ -17,6 +17,7 @@
> > >   *
> > >   */
> > >  
> > > +#include <clock.h>
> > >  #include <common.h>
> > >  #include <linux/sizes.h>
> > >  #include <memory.h>
> > > @@ -242,6 +243,10 @@ static int do_bootm_efi(struct image_data *data)
> > >  			       boot_header->ramdisk_image);
> > >  		printf("...\n");
> > >  	}
> > > +
> > > +	efi_set_variable_usec("LoaderTimeExecUSec", &efi_systemd_vendor_guid,
> > > +			      get_time_ns()/1000);
> > > +
> > >  	linux_efi_handover(handle, boot_header);
> > >  
> > >  	return 0;
> > > diff --git a/arch/efi/efi/efi.c b/arch/efi/efi/efi.c
> > > index b0e98f95b02b..a665a546be90 100644
> > > --- a/arch/efi/efi/efi.c
> > > +++ b/arch/efi/efi/efi.c
> > > @@ -340,6 +340,7 @@ efi_status_t efi_main(efi_handle_t image, efi_system_table_t *sys_table)
> > >  	efi_physical_addr_t mem;
> > >  	size_t memsize;
> > >  	efi_status_t efiret;
> > > +	char *uuid;
> > >  
> > >  #ifdef DEBUG
> > >  	sys_table->con_out->output_string(sys_table->con_out, L"barebox\n");
> > > @@ -377,6 +378,19 @@ efi_status_t efi_main(efi_handle_t image, efi_system_table_t *sys_table)
> > >  	mem_malloc_init((void *)mem, (void *)mem + memsize);
> > >  
> > >  	efi_clocksource_init();
> > > +	efi_set_variable_usec("LoaderTimeInitUSec", &efi_systemd_vendor_guid,
> > > +			      get_time_ns()/1000);
> > > +
> > > +	uuid = device_path_to_partuuid(device_path_from_handle(
> > > +				       efi_loaded_image->device_handle));
> > > +	if (uuid) {
> > > +		wchar_t *uuid16 = strdup_char_to_wchar(uuid);
> > > +		efi_set_variable("LoaderDevicePartUUID",
> > > +				 &efi_systemd_vendor_guid,
> > > +				 EFI_VARIABLE_BOOTSERVICE_ACCESS |
> > > +				 EFI_VARIABLE_RUNTIME_ACCESS,
> > > +				 uuid16, (strlen(uuid)+1) * sizeof(wchar_t));
> > > +	}
> > 
> > Shouldn't you free uuid here?
> 
> Yes, and uuid16.
> 
> Hmmm, what's the policy for OOM checks for string functions? None of the
> existing calls of strdup_char_to_wchar() check for NULL. And I didn't find
> any checks for asprintf() either.

Introducing xasprintf and using it throughout the tree is on my mental
todo list for some time now. I think You can skip the error check, or,
even better, introduce xasprintf ;)

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |