[PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'

Chao Liu posted 1 patch 5 months ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/cover.1749800810.git.chao.liu@yeah.net
Maintainers: "Dr. David Alan Gilbert" <dave@treblig.org>, Paolo Bonzini <pbonzini@redhat.com>, Peter Xu <peterx@redhat.com>, David Hildenbrand <david@redhat.com>, "Philippe Mathieu-Daudé" <philmd@linaro.org>
hmp-commands-info.hx    |   5 +-
include/system/memory.h |   3 +-
monitor/hmp-cmds.c      |   3 +-
system/memory.c         | 150 ++++++++++++++++++++++++++++++++++++----
4 files changed, 143 insertions(+), 18 deletions(-)
[PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'
Posted by Chao Liu 5 months ago
From: Chao Liu <chao.liu@yeah.net>

Hi, all:

After several rounds of discussion, I think that adding a -t option to the
`info mtree` command, which enables the display of tree-like node characters
(e.g., +--, |--), is a better approach.

As BALATON Zoltan pointed out, retaining space-based indentation for displaying
memory region (mr) nodes helps ensure that the output remains easily parseable
by other programs. This also provides better compatibility with existing tools
and scripts.

Users can choose between:

- Without '-t': The default format using spaces for indentation, suitable for
  scripting and parsing;
- With '-t': The tree-style format using visual characters, ideal for
  human-readable output.

This way, both usability and compatibility are preserved.

---

v7:
- Rebased on the top of current upstream:
  d9ce74873a6a ("Merge tag 'pull-vfio-20250611' of https://github.com/legoater/qemu
  into staging")
- Add -t option to 'info mtree' for tree-style node display

v6:
- Replaced the bool type with enum mtree_node_type to improve code readability;
- Modified the output to use only one horizontal dash instead of two, and
  aligned character printing for a cleaner look.

v5:
- Fix one comment, level 0 no longer prints line characters.

v4:
- When printing the child nodes of a single node, the line characters from the
  parent node's level are no longer printed, making the output more concise
  and clear overall;
- Use more commonly used ASCII characters, such as '|--' instead of '├──',
  and '`--' instead of '└──';
- Control the number of prefix characters to reduce unnecessary output.

v3:
- Use more maintainable c functions instead of macros, as per review comments.

v2:
- Enrich the commit message, add 'info mtree' print example;
- Optimize the code implementation according to the review comments.

--
Regards,
Chao

Chao Liu (1):
  system: improve visual representation of node hierarchy in 'info
    mtree' output for qemu monitor

 hmp-commands-info.hx    |   5 +-
 include/system/memory.h |   3 +-
 monitor/hmp-cmds.c      |   3 +-
 system/memory.c         | 150 ++++++++++++++++++++++++++++++++++++----
 4 files changed, 143 insertions(+), 18 deletions(-)

-- 
2.48.1


Re: [PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'
Posted by Markus Armbruster 5 months ago
Chao Liu <lc00631@tecorigin.com> writes:

> From: Chao Liu <chao.liu@yeah.net>
>
> Hi, all:
>
> After several rounds of discussion, I think that adding a -t option to the
> `info mtree` command, which enables the display of tree-like node characters
> (e.g., +--, |--), is a better approach.
>
> As BALATON Zoltan pointed out, retaining space-based indentation for displaying
> memory region (mr) nodes helps ensure that the output remains easily parseable
> by other programs. This also provides better compatibility with existing tools
> and scripts.

If people really feed the output of HMP info mtree to parsers, we should
probably provide the information via QMP.

> Users can choose between:
>
> - Without '-t': The default format using spaces for indentation, suitable for
>   scripting and parsing;
> - With '-t': The tree-style format using visual characters, ideal for
>   human-readable output.
>
> This way, both usability and compatibility are preserved.

Maintaining compatibility is a choice in HMP, not a requirement.  I'm
not judging, mind.
Re: [PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'
Posted by Chao Liu 5 months ago
On 2025/6/16 13:55, Markus Armbruster wrote:
> Chao Liu <lc00631@tecorigin.com> writes:
> 
>> From: Chao Liu <chao.liu@yeah.net>
>>
>> Hi, all:
>>
>> After several rounds of discussion, I think that adding a -t option to the
>> `info mtree` command, which enables the display of tree-like node characters
>> (e.g., +--, |--), is a better approach.
>>
>> As BALATON Zoltan pointed out, retaining space-based indentation for displaying
>> memory region (mr) nodes helps ensure that the output remains easily parseable
>> by other programs. This also provides better compatibility with existing tools
>> and scripts.
> 
> If people really feed the output of HMP info mtree to parsers, we should
> probably provide the information via QMP.
> 

Thank you for your helpful advice. I think the next step is to try 
implementing "info mtree" via QMP first, and then have it called by HMP.

I’ve added it to my to-do list, and I’ll try to implement it using QMP 
in the next phase.

>> Users can choose between:
>>
>> - Without '-t': The default format using spaces for indentation, suitable for
>>    scripting and parsing;
>> - With '-t': The tree-style format using visual characters, ideal for
>>    human-readable output.
>>
>> This way, both usability and compatibility are preserved.
> 
> Maintaining compatibility is a choice in HMP, not a requirement.  I'm
> not judging, mind.


Re: [PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'
Posted by Markus Armbruster 5 months ago
Chao Liu <chao.liu@yeah.net> writes:

> On 2025/6/16 13:55, Markus Armbruster wrote:
>> Chao Liu <lc00631@tecorigin.com> writes:
>> 
>>> From: Chao Liu <chao.liu@yeah.net>
>>>
>>> Hi, all:
>>>
>>> After several rounds of discussion, I think that adding a -t option to the
>>> `info mtree` command, which enables the display of tree-like node characters
>>> (e.g., +--, |--), is a better approach.
>>>
>>> As BALATON Zoltan pointed out, retaining space-based indentation for displaying
>>> memory region (mr) nodes helps ensure that the output remains easily parseable
>>> by other programs. This also provides better compatibility with existing tools
>>> and scripts.
>> 
>> If people really feed the output of HMP info mtree to parsers, we should
>> probably provide the information via QMP.
>
> Thank you for your helpful advice. I think the next step is to try implementing "info mtree" via QMP first, and then have it called by HMP.
>
> I’ve added it to my to-do list, and I’ll try to implement it using QMP in the next phase.

First question before you actually do that: use cases for feeding the
information to programs?  You might have answers already; I'm not on top
of prior conversations.

[...]
Re: [PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'
Posted by BALATON Zoltan 5 months ago
On Mon, 16 Jun 2025, Markus Armbruster wrote:
> Chao Liu <chao.liu@yeah.net> writes:
>
>> On 2025/6/16 13:55, Markus Armbruster wrote:
>>> Chao Liu <lc00631@tecorigin.com> writes:
>>>
>>>> From: Chao Liu <chao.liu@yeah.net>
>>>>
>>>> Hi, all:
>>>>
>>>> After several rounds of discussion, I think that adding a -t option to the
>>>> `info mtree` command, which enables the display of tree-like node characters
>>>> (e.g., +--, |--), is a better approach.
>>>>
>>>> As BALATON Zoltan pointed out, retaining space-based indentation for displaying
>>>> memory region (mr) nodes helps ensure that the output remains easily parseable
>>>> by other programs. This also provides better compatibility with existing tools
>>>> and scripts.
>>>
>>> If people really feed the output of HMP info mtree to parsers, we should
>>> probably provide the information via QMP.
>>
>> Thank you for your helpful advice. I think the next step is to try implementing "info mtree" via QMP first, and then have it called by HMP.
>>
>> I’ve added it to my to-do list, and I’ll try to implement it using QMP in the next phase.
>
> First question before you actually do that: use cases for feeding the
> information to programs?  You might have answers already; I'm not on top
> of prior conversations.

My request was to not make the output much wider than it is now as that 
would result in broken lines and less readable output. The comment was not 
about parsing output but keeping the result fit in not too wide terminals 
as 64 bit addresses makes it wide already. Thus replacing spaces with tree 
chars is OK with me as long as no new vertical space is added. The first 
version of patch increased vertical space. Using ASCII chars was request 
from somebody else but that makes the output look less nice so I'm not 
sure it worth the change with that. Maybe leaving current output then 
adding a separate tree mode with the original line drawing chars could 
satisfy all preferences.

Regards,
BALATON Zoltan
Re: [PATCH v7 0/1] Optimizing the print format of the QEMU monitor 'info mtree'
Posted by Chao Liu 5 months ago

On 2025/6/16 22:44, Markus Armbruster wrote:
> Chao Liu <chao.liu@yeah.net> writes:
> 
>> On 2025/6/16 13:55, Markus Armbruster wrote:
>>> Chao Liu <lc00631@tecorigin.com> writes:
>>>
>>>> From: Chao Liu <chao.liu@yeah.net>
>>>>
>>>> Hi, all:
>>>>
>>>> After several rounds of discussion, I think that adding a -t option to the
>>>> `info mtree` command, which enables the display of tree-like node characters
>>>> (e.g., +--, |--), is a better approach.
>>>>
>>>> As BALATON Zoltan pointed out, retaining space-based indentation for displaying
>>>> memory region (mr) nodes helps ensure that the output remains easily parseable
>>>> by other programs. This also provides better compatibility with existing tools
>>>> and scripts.
>>>
>>> If people really feed the output of HMP info mtree to parsers, we should
>>> probably provide the information via QMP.
>>
>> Thank you for your helpful advice. I think the next step is to try implementing "info mtree" via QMP first, and then have it called by HMP.
>>
>> I’ve added it to my to-do list, and I’ll try to implement it using QMP in the next phase.
> 
> First question before you actually do that: use cases for feeding the
> information to programs?  You might have answers already; I'm not on top
> of prior conversations.
> 
> [...]

The main purpose of this patch is not to make the memory-region tree 
output easier for programs to parse.

Instead, it's to make the output clearer when there are many memory 
regions in the machine.

Right now, we use spaces for indentation, which makes it hard to see the 
hierarchy clearly — I want similar to how the tree command works.

During discussions with multiple threads, we also looked into more 
details, like avoiding the use of special characters.