Re: [Qemu-devel] [qemu-s390x] [PATCH] s390-bios: Skip bootmap signature entries

[Christian Borntraeger <borntraeger@de.ibm.com>]

On 06.05.19 12:16, Thomas Huth wrote:
> On 06/05/2019 12.10, David Hildenbrand wrote:
>> On 06.05.19 12:01, David Hildenbrand wrote:
>>> On 29.04.19 15:09, Jason J. Herne wrote:
>>>> Newer versions of zipl have the ability to write signature entries to the boot
>>>> script for secure boot. We don't yet support secure boot, but we need to skip
>>>> over signature entries while reading the boot script in order to maintain our
>>>> ability to boot guest operating systems that have a secure bootloader.
>>>>
>>>> Signed-off-by: Jason J. Herne <jjherne@linux.ibm.com>
>>>> Reviewed-by: Farhan Ali <alifm@linux.ibm.com>
>>>> ---
>>>>  pc-bios/s390-ccw/bootmap.c | 19 +++++++++++++++++--
>>>>  pc-bios/s390-ccw/bootmap.h | 10 ++++++----
>>>>  2 files changed, 23 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/pc-bios/s390-ccw/bootmap.c b/pc-bios/s390-ccw/bootmap.c
>>>> index 7aef65a..d13b7cb 100644
>>>> --- a/pc-bios/s390-ccw/bootmap.c
>>>> +++ b/pc-bios/s390-ccw/bootmap.c
>>>> @@ -254,7 +254,14 @@ static void run_eckd_boot_script(block_number_t bmt_block_nr,
>>>>      memset(sec, FREE_SPACE_FILLER, sizeof(sec));
>>>>      read_block(block_nr, sec, "Cannot read Boot Map Script");
>>>>  
>>>> -    for (i = 0; bms->entry[i].type == BOOT_SCRIPT_LOAD; i++) {
>>>> +    for (i = 0; bms->entry[i].type == BOOT_SCRIPT_LOAD ||
>>>> +                bms->entry[i].type == BOOT_SCRIPT_SIGNATURE; i++) {
>>>> +
>>>> +        /* We don't support secure boot yet, so we skip signature entries */
>>>> +        if (bms->entry[i].type == BOOT_SCRIPT_SIGNATURE) {
>>>> +            continue;
>>>> +        }
>>>> +
>>>>          address = bms->entry[i].address.load_address;
>>>>          block_nr = eckd_block_num(&bms->entry[i].blkptr.xeckd.bptr.chs);
>>>>  
>>>> @@ -489,7 +496,15 @@ static void zipl_run(ScsiBlockPtr *pte)
>>>>  
>>>>      /* Load image(s) into RAM */
>>>>      entry = (ComponentEntry *)(&header[1]);
>>>> -    while (entry->component_type == ZIPL_COMP_ENTRY_LOAD) {
>>>> +    while (entry->component_type == ZIPL_COMP_ENTRY_LOAD ||
>>>> +           entry->component_type == ZIPL_COMP_ENTRY_SIGNATURE) {
>>>> +
>>>> +        /* We don't support secure boot yet, so we skip signature entries */
>>>> +        if (entry->component_type == ZIPL_COMP_ENTRY_SIGNATURE) {
>>>> +            entry++;
>>>> +            continue;
>>>> +        }
>>>> +
>>>>          zipl_load_segment(entry);
>>>>  
>>>>          entry++;
>>>> diff --git a/pc-bios/s390-ccw/bootmap.h b/pc-bios/s390-ccw/bootmap.h
>>>> index a085212..94f53a5 100644
>>>> --- a/pc-bios/s390-ccw/bootmap.h
>>>> +++ b/pc-bios/s390-ccw/bootmap.h
>>>> @@ -98,8 +98,9 @@ typedef struct ScsiMbr {
>>>>  #define ZIPL_COMP_HEADER_IPL    0x00
>>>>  #define ZIPL_COMP_HEADER_DUMP   0x01
>>>>  
>>>> -#define ZIPL_COMP_ENTRY_LOAD    0x02
>>>> -#define ZIPL_COMP_ENTRY_EXEC    0x01
>>>> +#define ZIPL_COMP_ENTRY_EXEC      0x01
>>>> +#define ZIPL_COMP_ENTRY_LOAD      0x02
>>>> +#define ZIPL_COMP_ENTRY_SIGNATURE 0x03
>>>>  
>>>>  typedef struct XEckdMbr {
>>>>      uint8_t magic[4];   /* == "xIPL"        */
>>>> @@ -117,8 +118,9 @@ typedef struct BootMapScriptEntry {
>>>>      BootMapPointer blkptr;
>>>>      uint8_t pad[7];
>>>>      uint8_t type;   /* == BOOT_SCRIPT_* */
>>>> -#define BOOT_SCRIPT_EXEC 0x01
>>>> -#define BOOT_SCRIPT_LOAD 0x02
>>>> +#define BOOT_SCRIPT_EXEC      0x01
>>>> +#define BOOT_SCRIPT_LOAD      0x02
>>>> +#define BOOT_SCRIPT_SIGNATURE 0x03
>>>>      union {
>>>>          uint64_t load_address;
>>>>          uint64_t load_psw;
>>>>
>>>
>>> Naive question from me:
>>>
>>> Can't we place the signatures somewhere else, and instead associate them
>>> with entries? This avoids breaking backwards compatibility for the sake
>>> of signatures we want unmodified zipl loaders to ignore.
>>>
>>
>>
>> ... but I guess this is already documented somewhere internally and
>> other components have been adjusted. IOW, cannot be changed anymore.
>>
>> Guess our implementation should have tolerated other entries than
>> "BOOT_SCRIPT_LOAD" right from the beginning.
> 
> Hmm, now we only tolerate the _LOAD and _SIGNATURE entries, but still
> nothing else... would it make sense to rewrite the code a little bit to
> tolerate all other kind of entries, but just act on the well-known _LOAD
> entries, so that we do not step into this trap in the future anymore?

I think we should not. Those entries might have sematic elements that the guest
wants to enforce. I do not think that this will come, but imagine a boot entry
that mandates some security wishes (e.g. do only run on non-shared cores).