Re: [Xen-devel] [PATCH v2] xen/arm: add warning if memory modules overlap

[Julien Grall <julien.grall.oss@gmail.com>]

On Thu, 17 Oct 2019 at 22:20, Brian Woods <brian.woods@xilinx.com> wrote:
>
> On Thu, Oct 17, 2019 at 09:34:51PM +0100, Julien Grall wrote:
> > Hi,
> >
> > On Thu, 17 Oct 2019 at 21:08, Brian Woods <brian.woods@xilinx.com> wrote:
> > >
> > > It's possible for a misconfigured device tree to cause Xen to crash when
> > > there are overlapping addresses in the memory modules.  Add a warning
> > > when printing the addresses to let the user know there's a possible
> > > issue.
> > >
> > > Signed-off-by: Brian Woods <brian.woods@xilinx.com>
> > > ---
> > > v1 -> v2
> > >         - removed nested loop and placed check in add_boot_module()
> > >
> > > Sample output:
> > > ...
> > > (XEN) MODULE[0]: 0000000001400000 - 0000000001542121 Xen
> > > (XEN) MODULE[1]: 0000000003846000 - 0000000003850080 Device Tree
> > > (XEN) MODULE[2]: 0000000003853000 - 0000000007fff676 Ramdisk
> > > (XEN) MODULE[3]: 0000000000080000 - 0000000003180000 Kernel
> > > (XEN)  RESVD[0]: 0000000003846000 - 0000000003850000
> > > (XEN)  RESVD[1]: 0000000003853000 - 0000000007fff676
> > > (XEN)
> > > (XEN) WARNING: overlap detected in the memory module addresses
> > > (XEN)
> > > (XEN) Command line: console=dtuart dtuart=serial0 dom0_mem=1G bootscrub=0 maxcpus=1 timer_slop=0
> > > ...
> > >
> > >  xen/arch/arm/bootfdt.c      | 4 ++++
> > >  xen/arch/arm/setup.c        | 6 ++++++
> > >  xen/include/asm-arm/setup.h | 1 +
> > >  3 files changed, 11 insertions(+)
> > >
> > > diff --git a/xen/arch/arm/bootfdt.c b/xen/arch/arm/bootfdt.c
> > > index 08fb59f..f8b34d4 100644
> > > --- a/xen/arch/arm/bootfdt.c
> > > +++ b/xen/arch/arm/bootfdt.c
> > > @@ -387,6 +387,10 @@ static void __init early_print_info(void)
> > >                 mem_resv->bank[j].start + mem_resv->bank[j].size - 1);
> > >      }
> > >      printk("\n");
> > > +
> > > +    if ( mem_module_overlap )
> > > +        printk("WARNING: overlap detected in the memory module addresses.\n");
> >
> > As a user such message would likely put me off. You tell me there are
> > an overlap, but you don't provide more information even if you likely
> > have the information in place. However...
>
> Well, I suppose the message could be changed to something like:
> "WARNING: overlap detected in the above memory module addresses."
> or something to more directly guide the users to the section.  Maybe
> move the 'printk("\n");' after the warning so it's grouped tighter with
> the module information.

My point stands even for this sort of message. You know the exact
overlap, so why would you hide it from the users?

>
> > > +
> > >      for ( i = 0 ; i < cmds->nr_mods; i++ )
> > >          printk("CMDLINE[%"PRIpaddr"]:%s %s\n", cmds->cmdline[i].start,
> > >                 cmds->cmdline[i].dt_name,
> > > diff --git a/xen/arch/arm/setup.c b/xen/arch/arm/setup.c
> > > index 705a917..315a131 100644
> > > --- a/xen/arch/arm/setup.c
> > > +++ b/xen/arch/arm/setup.c
> > > @@ -69,6 +69,8 @@ integer_param("xenheap_megabytes", opt_xenheap_megabytes);
> > >
> > >  domid_t __read_mostly max_init_domid;
> > >
> > > +bool __initdata mem_module_overlap;
> > > +
> > >  static __used void init_done(void)
> > >  {
> > >      /* Must be done past setting system_state. */
> > > @@ -254,6 +256,10 @@ struct bootmodule __init *add_boot_module(bootmodule_kind kind,
> > >                  mod->domU = false;
> > >              return mod;
> > >          }
> > > +
> > > +        if ( ((mod->start >= start) && (mod->start < start + size)) ||
> > > +             ((start >= mod->start) && (start < mod->start + mod->size)) )
> > > +            mem_module_overlap = true;
> >
> > ... What's wrong with just dumping the information here directly?
>
> IMO, it is better to have all the information printed in one spot.
> There is less to go through and easier to find out what is happening.
> There is also the fact that we do not have to print things twice (2 sets
> of names, starting addresses and ending addresses per overlap) when it
> is going to be printed in the near future anyway.  The cost of this is
> just one initdata bool, which while I am not thrilled about, does not
> seem that expensive (compared to a nested loop or printing out at least
> (16*2 + 12) * 2 characters per overlap(at least on Arm64)).

Again, this is boot code and not a path that is going to be called
hundreds of time. So performance is the last thing I care in this
patch.

If we try to help the users by telling them there is an overlap
between modules, then we should do it properly and tell them the exact
overlap. Otherwise this is nearly as pointless as a crash later on in
the boot process.

I also don't want a double for loop or any additional global variable
when it can be done by simply adding a check in add_boot_module().

Cheers,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel