[PATCH] btrfs-progs: device: print num_stripes in usage command

[PATCH] btrfs-progs: device: print num_stripes in usage command

[Sidong Yang]

This patch appends num_stripes for each chunks in device usage commands.
It helps to see profiles easily. The output is like below.

/dev/vdb, ID: 1
   Device size:             1.00GiB
   Device slack:              0.00B
   Data,single[1]:        120.00MiB
   Metadata,DUP[2]:       102.38MiB
   System,DUP[2]:          16.00MiB
   Unallocated:           785.62MiB

Signed-off-by: Sidong Yang <realwakka@gmail.com>
---
 cmds/filesystem-usage.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/cmds/filesystem-usage.c b/cmds/filesystem-usage.c
index 107453d4..4d8c0c2d 100644
--- a/cmds/filesystem-usage.c
+++ b/cmds/filesystem-usage.c
@@ -1192,6 +1192,7 @@ void print_device_chunks(struct device_info *devinfo,
 		const char *r_mode;
 		u64 flags;
 		u64 size;
+		u64 num_stripes;
 
 		if (chunks_info_ptr[i].devid != devinfo->devid)
 			continue;
@@ -1201,10 +1202,13 @@ void print_device_chunks(struct device_info *devinfo,
 		description = btrfs_group_type_str(flags);
 		r_mode = btrfs_group_profile_str(flags);
 		size = calc_chunk_size(chunks_info_ptr+i);
-		printf("   %s,%s:%*s%10s\n",
+		num_stripes = chunks_info_ptr[i].num_stripes;
+		printf("   %s,%s[%llu]:%*s%10s\n",
 			description,
 			r_mode,
-			(int)(20 - strlen(description) - strlen(r_mode)), "",
+			num_stripes,
+			(int)(20 - strlen(description) - strlen(r_mode)
+				  - count_digits(num_stripes) - 2), "",
 			pretty_size_mode(size, unit_mode));
 
 		allocated += size;
-- 
2.25.1

Re: [PATCH] btrfs-progs: device: print num_stripes in usage command

[David Sterba]

On Sun, May 30, 2021 at 12:56:36PM +0000, Sidong Yang wrote:
> This patch appends num_stripes for each chunks in device usage commands.
> It helps to see profiles easily. The output is like below.
> 
> /dev/vdb, ID: 1
>    Device size:             1.00GiB
>    Device slack:              0.00B
>    Data,single[1]:        120.00MiB
>    Metadata,DUP[2]:       102.38MiB
>    System,DUP[2]:          16.00MiB
>    Unallocated:           785.62MiB

This is example of single and dup, but the whole point was to print that
for the striped profiles, like is shown in the issue
https://github.com/kdave/btrfs-progs/issues/372 , so it needs to be
conditional.

As the mirrored profiles have a fixed number of copies, there's no doubt
on how many devices are printed, same for single and dup.

I'm more inclined to use the "/" as separator: 'RAID6/3' it's more like
"raid6 on three devices", the "[]" looks like indexing. "," is used as
separator of the type so it won't be as simple to parse that.

Re: [PATCH] btrfs-progs: device: print num_stripes in usage command

[Sidong Yang]

On Mon, Jun 07, 2021 at 09:20:49PM +0200, David Sterba wrote:
> On Sun, May 30, 2021 at 12:56:36PM +0000, Sidong Yang wrote:
> > This patch appends num_stripes for each chunks in device usage commands.
> > It helps to see profiles easily. The output is like below.
> > 
> > /dev/vdb, ID: 1
> >    Device size:             1.00GiB
> >    Device slack:              0.00B
> >    Data,single[1]:        120.00MiB
> >    Metadata,DUP[2]:       102.38MiB
> >    System,DUP[2]:          16.00MiB
> >    Unallocated:           785.62MiB
> 
Hi David. Thanks for review!

> This is example of single and dup, but the whole point was to print that
> for the striped profiles, like is shown in the issue
> https://github.com/kdave/btrfs-progs/issues/372 , so it needs to be
> conditional.

Okay, striped profiles need to be printed. I read some page from
wikipedia about raid. and I got that RAID 0,5,6,10 supports striping.
Those need to be printed and no printing for RAID 1, single, dup.

> 
> As the mirrored profiles have a fixed number of copies, there's no doubt
> on how many devices are printed, same for single and dup.
> 
> I'm more inclined to use the "/" as separator: 'RAID6/3' it's more like
> "raid6 on three devices", the "[]" looks like indexing. "," is used as
> separator of the type so it won't be as simple to parse that.

I agree. I'll use "/" for next version.