[U-Boot] [PATCH 1/1] x86: show UEFI images involved in crash

[Heinrich Schuchardt]

On 8/27/19 3:18 AM, Bin Meng wrote:
> Hi Heinrich,
>
> On Tue, Aug 27, 2019 at 1:26 AM Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>
>> On 8/26/19 8:13 AM, Bin Meng wrote:
>>> Hi Heinrich,
>>>
>>> On Mon, Aug 26, 2019 at 1:55 AM Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>>>
>>>> If a crash occurs, show the loaded UEFI images to facilitate analysis.
>>>>
>>>> This is an example output:
>>>>
>>>> => bootefi 0x1000000
>>>> Found 0 disks
>>>> Hello world of bugs!
>>>> Invalid Opcode (Undefined Opcode)
>>>> EIP: 0010:[<06ceb06e>] EFLAGS: 00010206
>>>> Original EIP :[<fec9906e>]
>>>> EAX: 00000000 EBX: 06cec000 ECX: 00000fd0 EDX: 00000001
>>>> ESI: 06ced18a EDI: 07d0fe10 EBP: 07fe27a0 ESP: 07d0fde0
>>>>    DS: 0018 ES: 0018 FS: 0020 GS: 0018 SS: 0018
>>>> CR0: 00000033 CR2: 00000000 CR3: 00000000 CR4: 00000000
>>>> DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
>>>> DR6: ffff0ff0 DR7: 00000400
>>>> Stack:
>>>>       0x07d0fde8 : 0x00000000
>>>>       0x07d0fde4 : 0x06ced040
>>>> --->0x07d0fde0 : 0x07fe27a0
>>>>       0x07d0fddc : 0x00010206
>>>>       0x07d0fdd8 : 0x00000010
>>>>       0x07d0fdd4 : 0x06ceb06e
>>>> UEFI image [0x06cea000:0x06cf0fff] pc=0x106e '/bug-i386.efi'
>>>> ### ERROR ### Please RESET the board ###
>>>>
>>>> With the additional information provided by this patch we know that the
>>>> problem occurred 0x106e after the load address of bug-i386.efi.
>>>>
>>>> Signed-off-by: Heinrich Schuchardt <xypron.glpk at gmx.de>
>>>> ---
>>>>    arch/x86/cpu/i386/interrupt.c | 14 ++++++++++++++
>>>>    1 file changed, 14 insertions(+)
>>>>
>>>> diff --git a/arch/x86/cpu/i386/interrupt.c b/arch/x86/cpu/i386/interrupt.c
>>>> index 47df3172b7..1445204878 100644
>>>> --- a/arch/x86/cpu/i386/interrupt.c
>>>> +++ b/arch/x86/cpu/i386/interrupt.c
>>>> @@ -12,6 +12,7 @@
>>>>
>>>>    #include <common.h>
>>>>    #include <dm.h>
>>>> +#include <efi_loader.h>
>>>>    #include <asm/control_regs.h>
>>>>    #include <asm/i8259.h>
>>>>    #include <asm/interrupt.h>
>>>> @@ -64,6 +65,18 @@ static char *exceptions[] = {
>>>>           "Reserved"
>>>>    };
>>>>
>>>> +/**
>>>> + * show_efi_loaded_images() - show loaded UEFI images
>>>> + *
>>>> + * List all loaded UEFI images.
>>>> + *
>>>> + * @eip:       instruction pointer
>>>> + */
>>>> +static void show_efi_loaded_images(uintptr_t eip)
>>>> +{
>>>> +       efi_print_image_infos((void *)eip);
>>>> +}
>>>> +
>>>>    static void dump_regs(struct irq_regs *regs)
>>>>    {
>>>>           unsigned long cs, eip, eflags;
>>>> @@ -144,6 +157,7 @@ static void dump_regs(struct irq_regs *regs)
>>>>                   printf("0x%8.8lx : 0x%8.8lx\n", sp, (ulong)readl(sp));
>>>>                   sp -= 4;
>>>>           }
>>>> +       show_efi_loaded_images(eip);
>>>
>>> Should we wrap the call with #ifdef CONFIG_EFI_LOADER or something?
>>
>> In include/efi_loader.h we have
>>
>> static inline void efi_print_image_infos(void *pc) { }
>>
>> if EFI_LOADER is not defined.
>>
>> Best regards
>>
>
> I feel a little bit strange of show_efi_loaded_images() being called
> in the dump_regs(). The dump_regs() is called in the exception
> handler. It's a bit odd we show the EFI image info in the exception
> handler. Shouldn't we print the EFI image info from the command line
> interface?

We have a command to  display loaded images from the command line
(efidebug images). But when a crash occurs in an UEFI application, I
cannot reach the command line anymore.

When running complex UEFI applications like the SCT multiple UEFI images
are involved with loading addresses which I cannot control from the
command line. So it is hard to find out where a crash occurred.

For ARM we already have the same information in crash dumps.

If not UEFI image is loaded, no line is added to the crash dump.

Best regards

Heinrich