On 03.12.17 22:20, xypron.glpk@gmx.de wrote:
On Sunday, December 3, 2017 8:24:39 PM CET Mark Kettenis wrote:
Heinrich Schuchardt schreef op 2017-12-02 13:42:
For each block device we have to setup:
a device path for the block device
a partition device path for the block device with
partition number 0
Do x86 UEFI implementations actually generate these partition number 0 paths?
Thanks for reviewing.
I do not have access the an x86 computer with UEFI firmware. Maybe somebody else on the list has.
Another reference point is EDK2.
Part of the relevant coding is in function PartitionInstallMbrChildHandles(), MdeModulePkg/Universal/Disk/PartitionDxe/Mbr.c:47.
Another part is in PartitionInstallChildHandle(), MdeModulePkg/Universal/Disk/PartitionDxe/Partition.c.
I could not identify any line referring to partition 0.
The standard says that 0 can be used in tis fashion but doesn't seem to prescribe their use.
<cite>A partition number of zero can be used to represent the raw hard drive or a raw extended partition.</cite>
Do you think we should not generate this entry?
Given that basically all major UEFI implementations are based on edk2, I think we should just do it exactly the same way :)
https://wiki.linaro.org/LEG/UEFIforQEMU
To give you example grub lsefi output for a simple disk with 2 MBR partitions:
== U-Boot w/o this patch set (SD) ==
/HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/EndEntire block device path Handle 0x9fef8f78 /HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/HD(0,800,32000,2de808cb00000000,1,1)/EndEntire block device path simple FS Handle 0x9fef90a0 /HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/HD(1,32800,1cd800,2de808cb00000000,1,1)/EndEntire block device path simple FS
== U-Boot w/ this patch set (SD) ==
Handle 0x9fef7598 /HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/EndEntire block device path simple FS Handle 0x9fef76c0 /HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/HD(0,0,200000,2de808cb00000000,1,1)/EndEntire block device path Handle 0x9fef7020 /HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/HD(1,800,32000,2de808cb00000000,1,1)/EndEntire block device path simple FS Handle 0x9fef7228 /HardwareVendor(e61d73b9-a384-4acc-aeab-82e828f3628b)[0: ]/USB(6,0)/HD(2,32800,1cd800,2de808cb00000000,1,1)/EndEntire block device path simple FS
== edk2 (virtio) ==
Handle 0x43250d10 /ACPI(a0341d0,0)/PCI(0,1)/EndEntire disk block fa920010-6785-4941-b6ec-498c579f160a PCI device path Handle 0x4223af90 /ACPI(a0341d0,0)/PCI(0,1)/HD(1,800,32000,2de808cb00000000,1,1)/EndEntire simple FS disk c12a7328-f81f-11d2-ba4b-00a0c93ec93b 8cf2f62c-bc9b-4821-808d-ec9ec421a1a0 block device path Handle 0x4223ac90
/ACPI(a0341d0,0)/PCI(0,1)/HD(2,32800,1cd800,2de808cb00000000,1,1)/EndEntire simple FS disk 8cf2f62c-bc9b-4821-808d-ec9ec421a1a0 block device path
=====
So I think the only patch we really need is something like this:
diff --git a/lib/efi_loader/efi_device_path.c b/lib/efi_loader/efi_device_path.c index b4e2f933cb..9ba6b04e5e 100644 --- a/lib/efi_loader/efi_device_path.c +++ b/lib/efi_loader/efi_device_path.c @@ -416,7 +416,7 @@ static void *dp_part_fill(void *buf, struct blk_desc *desc, int part) hddp->dp.type = DEVICE_PATH_TYPE_MEDIA_DEVICE; hddp->dp.sub_type = DEVICE_PATH_SUB_TYPE_HARD_DRIVE_PATH; hddp->dp.length = sizeof(*hddp); - hddp->partition_number = part - 1; + hddp->partition_number = part; hddp->partition_start = info.start; hddp->partition_end = info.size; if (desc->part_type == PART_TYPE_EFI)
But this can wait until after rc1, so I'll send the pull request as is out now :).
Alex