[ipxe-devel] Bug in UEFI Sanboot with iso9660

Mon Dec 12 21:57:59 UTC 2016

The windows installer ISO I tested which uefi sanboots both with and without the blksize_shift in place seems to be exactly this structure (just an iso filesystem). I have an ol7 installer iso which has an MBR structure which also boots with and without the shift. I've yet to find an iso that will not boot with the blksize_shift code removed. If anyone has any others they would like me to try, please link me to them.

Vish

> On Dec 12, 2016, at 11:42 AM, Christian Nilsson <nikize at gmail.com> wrote:
> 
> I think it is also "valid" (as in common practice to implement support) to have a iso9660 and search for /EFI/BOOT/BOOTX64.EFI or /EFI/BOOT/BOOTIA32.EFI depending on architechture.
> 
> So maybe it should search for GPT structure first and if it exists skip possible 9660 detection?
> 
> On Mon, Dec 12, 2016 at 8:23 PM, Vish Ishaya Abrams <vish.ishaya at oracle.com> wrote:
> Hello
> 
> With current master, after installing windows 2K12 onto an iscsi drive and trying to sanboot in UEFI mode it fails. I tracked the issue down to the iso9660 detection code. It appears that a windows 2K12 install appears as an iso9660 installation in addition to having a GPT:
> 
> # isoinfo -d -i /dev/sdb8
> CD-ROM is in ISO 9660 format
> System id:
> Volume id: IR3_SSS_X64FREE_EN-US_DV9
> Volume set id: IR3_SSS_X64FREE_EN-US_DV9
> Publisher id: MICROSOFT CORPORATION
> Data preparer id: MICROSOFT CORPORATION, ONE MICROSOFT WAY, REDMOND WA 98052, (425) 882-8080
> Application id: CDIMAGE 2.53 (01/01/2005 TM)
> Copyright File id:
> Abstract File id:
> Bibliographic File id:
> Volume set size is: 1
> Volume set sequence number is: 1
> Logical block size is: 2048
> Volume size is: 2217916
> El Torito VD version 1 found, boot catalog is in sector 22
> NO Joliet present
> NO Rock Ridge present
> Eltorito validation header:
>     Hid 1
>     Arch 0 (x86)
>     ID 'Microsoft Corporation'
>     Key 55 AA
>     Eltorito defaultboot header:
>         Bootid 88 (bootable)
>         Boot media 0 (No Emulation Boot)
>         Load segment 0
>         Sys type 0
>         Nsect 8
>         Bootoff 876 2166
> 
> # fdisk --list /dev/sdb8
> 
> Disk /dev/sdb8: 18.6 GiB, 20004052992 bytes, 39070416 sectors
> Units: sectors of 1 * 512 = 512 bytes
> Sector size (logical/physical): 512 bytes / 4096 bytes
> I/O size (minimum/optimal): 4096 bytes / 4096 bytes
> Disklabel type: gpt
> Disk identifier: 8071B2D8-D75D-49E3-B3DD-B0401535345E
> 
> Device        Start      End  Sectors  Size Type
> /dev/sdb8p1    2048   616447   614400  300M Windows recovery environment
> /dev/sdb8p2  616448   819199   202752   99M EFI System
> /dev/sdb8p3  819200  1081343   262144  128M Microsoft reserved
> /dev/sdb8p4 1081344 39069695 37988352 18.1G Microsoft basic data
> 
> # file -s /dev/sdb8
> /dev/sdb8: DOS/MBR boot sector MS-MBR Windows 7 english at offset 0x163 "Invalid partition table" at offset 0x17b "Error loading operating system" at offset 0x19a "Missing operating system" ISO 9660 CD-ROM filesystem data 'IR3_SSS_X64FREE_EN-US_DV9' (bootable); partition 1 : ID=0xee, start-CHS (0x0,0,2), end-CHS (0x3ff,255,63), startsector 1, 4294967295 sectors
> 
> the efi_block code detects this as an iso_9660 filesystem and sets a blksize_shift of 2 and it does not successfully read the EFI filesystem. If I comment out the blksize_shift setting, it boots just fine. I also tested booting a couple of actual iso filesystems which detected a blksize_shift of 2. Both the ubuntu installer iso and the windows installer iso boot just fine with the blksize_shift commented out.
> 
> This leads me to believe that the blksize_shift code is not necessary in uefi mode. I have commented it out in my build, but I'm not sure if there are some other odd iso images which require this setting. If so, then we need a way to detect the above case where we have an iso9660 that also has an EFI GPT and we can disable setting the blksize_shift in that case.
> 
> Any thoughts?
> Vish
> 
> 
> _______________________________________________
> ipxe-devel mailing list
> ipxe-devel at lists.ipxe.org
> https://lists.ipxe.org/mailman/listinfo.cgi/ipxe-devel
> 