Re: [PULL SUBSYSTEM qemu-pseries] pseries: Update SLOF firmware image

[Cédric Le Goater]

On 2/20/20 2:50 AM, Alexey Kardashevskiy wrote:
> 
> 
> On 19/02/2020 18:18, Cédric Le Goater wrote:
>> On 2/19/20 7:44 AM, Alexey Kardashevskiy wrote:
>>>
>>>
>>> On 19/02/2020 12:20, Alexey Kardashevskiy wrote:
>>>>
>>>>
>>>> On 18/02/2020 23:59, Cédric Le Goater wrote:
>>>>> On 2/18/20 1:48 PM, Cédric Le Goater wrote:
>>>>>> On 2/18/20 10:40 AM, Cédric Le Goater wrote:
>>>>>>> On 2/18/20 10:10 AM, Alexey Kardashevskiy wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>> On 18/02/2020 20:05, Alexey Kardashevskiy wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 18/02/2020 18:12, Cédric Le Goater wrote:
>>>>>>>>>> On 2/18/20 1:30 AM, Alexey Kardashevskiy wrote:
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 17/02/2020 20:48, Cédric Le Goater wrote:
>>>>>>>>>>>> On 2/17/20 3:12 AM, Alexey Kardashevskiy wrote:
>>>>>>>>>>>>> The following changes since commit 
>>>>>>>>>>>>> 05943fb4ca41f626078014c0327781815c6584c5:
>>>>>>>>>>>>>
>>>>>>>>>>>>>   ppc: free 'fdt' after reset the machine (2020-02-17 11:27:23 
>>>>>>>>>>>>> +1100)
>>>>>>>>>>>>>
>>>>>>>>>>>>> are available in the Git repository at:
>>>>>>>>>>>>>
>>>>>>>>>>>>>   address@hidden:aik/qemu.git tags/qemu-slof-20200217
>>>>>>>>>>>>>
>>>>>>>>>>>>> for you to fetch changes up to 
>>>>>>>>>>>>> ea9a03e5aa023c5391bab5259898475d0298aac2:
>>>>>>>>>>>>>
>>>>>>>>>>>>>   pseries: Update SLOF firmware image (2020-02-17 13:08:59 +1100)
>>>>>>>>>>>>>
>>>>>>>>>>>>> ----------------------------------------------------------------
>>>>>>>>>>>>> Alexey Kardashevskiy (1):
>>>>>>>>>>>>>       pseries: Update SLOF firmware image
>>>>>>>>>>>>>
>>>>>>>>>>>>>  pc-bios/README   |   2 +-
>>>>>>>>>>>>>  pc-bios/slof.bin | Bin 931032 -> 968560 bytes
>>>>>>>>>>>>>  roms/SLOF        |   2 +-
>>>>>>>>>>>>>  3 files changed, 2 insertions(+), 2 deletions(-)
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> *** Note: this is not for master, this is for pseries
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Hello Alexey,
>>>>>>>>>>>>
>>>>>>>>>>>> QEMU fails to boot from disk. See below.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> It does boot mine (fedora 30, ubuntu 18.04), see below. I believe I
>>>>>>>>>>> could have broken something but I need more detail. Thanks,
>>>>>>>>>>
>>>>>>>>>> fedora31 boots but not ubuntu 19.10. Could it be GRUB version 2.04 ? 
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> No, not that either:
>>>>>>>>
>>>>>>>>
>>>>>>>> but it might be because of power9 - I only tried power8, rsyncing the
>>>>>>>> image to a p9 machine now...
>>>>>>>
>>>>>>> Here is the disk : 
>>>>>>>
>>>>>>> Disk /dev/sda: 50 GiB, 53687091200 bytes, 104857600 sectors
>>>>>>> Disk model: QEMU HARDDISK   
>>>>>>> Units: sectors of 1 * 512 = 512 bytes
>>>>>>> Sector size (logical/physical): 512 bytes / 512 bytes
>>>>>>> I/O size (minimum/optimal): 512 bytes / 512 bytes
>>>>>>> Disklabel type: gpt
>>>>>>> Disk identifier: 27DCE458-231A-4981-9FF1-983F87C2902D
>>>>>>>
>>>>>>> Device         Start       End   Sectors Size Type
>>>>>>> /dev/sda1       2048     16383     14336   7M PowerPC PReP boot
>>>>>>> /dev/sda2      16384 100679679 100663296  48G Linux filesystem
>>>>>>> /dev/sda3  100679680 104857566   4177887   2G Linux swap
>>>>>>>
>>>>>>>
>>>>>>> GPT ? 
>>>>>>
>>>>>> For the failure, I bisected up to :
>>>>>>
>>>>>> f12149908705 ("ext2: Read all 64bit of inode number")
>>>>>
>>>>> Here is a possible fix for it. I did some RPN on my hp28s in the past 
>>>>> but I am not forth fluent.
>>>>
>>>>
>>>> you basically zeroed the top bits by shifting them too far right :)
>>>>
>>>> The proper fix I think is:
>>>>
>>>> -  32 lshift or
>>>> +  20 lshift or
>>>>
>>>> I keep forgetting it is all in hex. Can you please give it a try? My
>>>> 128GB disk does not expose this problem somehow. Thanks,
>>>
>>> Better try this one please:
>>>
>>> https://github.com/aik/SLOF/tree/ext4
>> Tested with the same image. Looks good. 
> 
> 
> Thanks for testing. But it is still bizarre behaviour, why do we end up
> there anyway...
> 
> 
>>> What I still do not understand is why GRUB is using ext2 from SLOF, it
>>> should parse ext4 itself :-/
>>
>> Here is the fs information.
>>
>>
>> Filesystem volume name:   <none>
>> Last mounted on:          /
>> Filesystem UUID:          8d53f6b4-ffc2-4d8f-bd09-67ac97d7b0c5
>> Filesystem magic number:  0xEF53
>> Filesystem revision #:    1 (dynamic)
>> Filesystem features:      has_journal ext_attr resize_inode dir_index 
>> filetype needs_recovery extent flex_bg sparse_super large_file huge_file 
>> uninit_bg dir_nlink extra_isize
> 
> 
> huh, this one does not have 64bit like mine, I blindly assumed that by
> 2020 everything would be using that. Well that explains the bug. And
> yours also has uninit_bg (the whole idea of this flag is not obvious but
> ok).
> 
> 
>> Filesystem flags:         unsigned_directory_hash 
>> Default mount options:    user_xattr acl
>> Filesystem state:         clean
>> Errors behavior:          Continue
>> Filesystem OS type:       Linux
>> Inode count:              3127296
>> Block count:              12582912
>> Reserved block count:     552210
>> Free blocks:              7907437
>> Free inodes:              2863361
>> First block:              0
>> Block size:               4096
>> Fragment size:            4096
> 
> 
> Mine here has:
> Group descriptor size:    64
> 
> so there is no "hi" part and this is what my fix now handles (0x32 vs.
> 0x20 was not the problem actually).
> 
> Did you do this on purpose or the installer did it? :)

I don't remember. I think I have kept the same disk image since 14.04 
and did some fs resize.

> Anyway, I'd like to get this particular disk image to understand why on
> earth it is using the ext2 package from SLOF. Thanks,

Sure.

Thanks,

C.

> 
>> Reserved GDT blocks:      1021
>> Blocks per group:         32768
>> Fragments per group:      32768
>> Inodes per group:         8144
>> Inode blocks per group:   509
>> Flex block group size:    16
>> Filesystem created:       Wed Dec 14 15:40:55 2016
>> Last mount time:          Wed Feb 19 08:06:52 2020
>> Last write time:          Wed Feb 19 08:06:46 2020
>> Mount count:              1863
>> Maximum mount count:      -1
>> Last checked:             Fri Nov 23 19:09:13 2018
>> Check interval:           0 (<none>)
>> Lifetime writes:          883 GB
>> Reserved blocks uid:      0 (user root)
>> Reserved blocks gid:      0 (group root)
>> First inode:              11
>> Inode size:            256
>> Required extra isize:     28
>> Desired extra isize:      28
>> Journal inode:            8
>> Default directory hash:   half_md4
>> Directory Hash Seed:      f7cb5863-4885-47b6-b24b-369df6a3b1a4
>> Journal backup:           inode blocks
>> Journal features:         journal_incompat_revoke
>> Journal size:             128M
>> Journal length:           32768
>> Journal sequence:         0x0004beb2
>>
>> Thanks,
>>
>> C.
>>
>>>
>>>>
>>>>
>>>>>
>>>>> "slash not found" is still there though. 
>>>
>>>
>>> Yeah I see these but they are harmless as far as I can tell.
>>>
>>>
>>>
>>>>>
>>>>> Cheers,
>>>>>
>>>>> C.
>>>>>
>>>>>
>>>>> From 92dc9f6dc7c6434419306d5a382adb42169b712a Mon Sep 17 00:00:00 2001
>>>>> From: =?UTF-8?q?C=C3=A9dric=20Le=20Goater?= <address@hidden>
>>>>> Date: Tue, 18 Feb 2020 13:54:54 +0100
>>>>> Subject: [PATCH] ext2: Fix 64bit inode number
>>>>> MIME-Version: 1.0
>>>>> Content-Type: text/plain; charset=UTF-8
>>>>> Content-Transfer-Encoding: 8bit
>>>>>
>>>>> Fixes: f12149908705 ("ext2: Read all 64bit of inode number")
>>>>> Signed-off-by: Cédric Le Goater <address@hidden>
>>>>> ---
>>>>>  slof/fs/packages/ext2-files.fs | 2 +-
>>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/slof/fs/packages/ext2-files.fs 
>>>>> b/slof/fs/packages/ext2-files.fs
>>>>> index b6a7880bd88e..f1d9fdfd67e2 100644
>>>>> --- a/slof/fs/packages/ext2-files.fs
>>>>> +++ b/slof/fs/packages/ext2-files.fs
>>>>> @@ -152,7 +152,7 @@ CONSTANT /ext4-ee
>>>>>    dup
>>>>>    8 + l@-le               \ reads bg_inode_table_lo
>>>>>    swap 28 + l@-le         \ reads bg_inode_table_hi
>>>>> -  32 lshift or
>>>>> +  32 rshift or
>>>>>    block-size @ *          \ # in group, inode table
>>>>>    swap inode-size @ * + xlsplit seek drop  inode @ inode-size @ read drop
>>>>>  ;
>>>>>
>>>>
>>>
>>
>