[PATCH] image_types: ext4 should always have large inodes

[PATCH] image_types: ext4 should always have large inodes

[Ross Burton]

mkfs.ext4 has logic where "small" file systems use "good old" 128-byte
inodes instead of the default 256-byte inodes.  This means that small
filesystems are not 2038-safe as they only support 32-bit timestamps.

However, "small" is defined as "less than 500MB", which is a fair number
of filesystems in OpenEmbedded, especially when filesystems may be
created small and expanded later.

To avoid a discussion over how small a file system should be before
it shouldn't be 2038-safe, simply build all filesystems with the default
type so they all have 256-byte inodes and thus 64-bit timestamps.

[ YOCTO #14478 ]

Signed-off-by: Ross Burton <ross.burton@arm.com>
---
 meta/classes/image_types.bbclass | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/classes/image_types.bbclass b/meta/classes/image_types.bbclass
index 954d6739ec..3dad86caa7 100644
--- a/meta/classes/image_types.bbclass
+++ b/meta/classes/image_types.bbclass
@@ -90,7 +90,7 @@ oe_mkext234fs () {
 
 IMAGE_CMD_ext2 = "oe_mkext234fs ext2 ${EXTRA_IMAGECMD}"
 IMAGE_CMD_ext3 = "oe_mkext234fs ext3 ${EXTRA_IMAGECMD}"
-IMAGE_CMD_ext4 = "oe_mkext234fs ext4 ${EXTRA_IMAGECMD}"
+IMAGE_CMD_ext4 = "oe_mkext234fs ext4 -T default ${EXTRA_IMAGECMD}"
 
 MIN_BTRFS_SIZE ?= "16384"
 IMAGE_CMD_btrfs () {
-- 
2.25.1

Re: [OE-core] [PATCH] image_types: ext4 should always have large inodes

[Alexandre Belloni]

Hi Ross,

This broke at least non-gpl3:
https://autobuilder.yoctoproject.org/typhoon/#/builders/75/builds/3691/steps/11/logs/stdio

On 16/07/2021 13:24:07+0100, Ross Burton wrote:
> mkfs.ext4 has logic where "small" file systems use "good old" 128-byte
> inodes instead of the default 256-byte inodes.  This means that small
> filesystems are not 2038-safe as they only support 32-bit timestamps.
> 
> However, "small" is defined as "less than 500MB", which is a fair number
> of filesystems in OpenEmbedded, especially when filesystems may be
> created small and expanded later.
> 
> To avoid a discussion over how small a file system should be before
> it shouldn't be 2038-safe, simply build all filesystems with the default
> type so they all have 256-byte inodes and thus 64-bit timestamps.
> 
> [ YOCTO #14478 ]
> 
> Signed-off-by: Ross Burton <ross.burton@arm.com>
> ---
>  meta/classes/image_types.bbclass | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/meta/classes/image_types.bbclass b/meta/classes/image_types.bbclass
> index 954d6739ec..3dad86caa7 100644
> --- a/meta/classes/image_types.bbclass
> +++ b/meta/classes/image_types.bbclass
> @@ -90,7 +90,7 @@ oe_mkext234fs () {
>  
>  IMAGE_CMD_ext2 = "oe_mkext234fs ext2 ${EXTRA_IMAGECMD}"
>  IMAGE_CMD_ext3 = "oe_mkext234fs ext3 ${EXTRA_IMAGECMD}"
> -IMAGE_CMD_ext4 = "oe_mkext234fs ext4 ${EXTRA_IMAGECMD}"
> +IMAGE_CMD_ext4 = "oe_mkext234fs ext4 -T default ${EXTRA_IMAGECMD}"
>  
>  MIN_BTRFS_SIZE ?= "16384"
>  IMAGE_CMD_btrfs () {
> -- 
> 2.25.1
> 

> 
> 
> 


-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Re: [OE-core] [PATCH] image_types: ext4 should always have large inodes

[Ross Burton]

Drop this, a second iteration is now on the list.

Ross

On Fri, 16 Jul 2021 at 15:26, Alexandre Belloni
<alexandre.belloni@bootlin.com> wrote:
>
> Hi Ross,
>
> This broke at least non-gpl3:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/75/builds/3691/steps/11/logs/stdio
>
> On 16/07/2021 13:24:07+0100, Ross Burton wrote:
> > mkfs.ext4 has logic where "small" file systems use "good old" 128-byte
> > inodes instead of the default 256-byte inodes.  This means that small
> > filesystems are not 2038-safe as they only support 32-bit timestamps.
> >
> > However, "small" is defined as "less than 500MB", which is a fair number
> > of filesystems in OpenEmbedded, especially when filesystems may be
> > created small and expanded later.
> >
> > To avoid a discussion over how small a file system should be before
> > it shouldn't be 2038-safe, simply build all filesystems with the default
> > type so they all have 256-byte inodes and thus 64-bit timestamps.
> >
> > [ YOCTO #14478 ]
> >
> > Signed-off-by: Ross Burton <ross.burton@arm.com>
> > ---
> >  meta/classes/image_types.bbclass | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/meta/classes/image_types.bbclass b/meta/classes/image_types.bbclass
> > index 954d6739ec..3dad86caa7 100644
> > --- a/meta/classes/image_types.bbclass
> > +++ b/meta/classes/image_types.bbclass
> > @@ -90,7 +90,7 @@ oe_mkext234fs () {
> >
> >  IMAGE_CMD_ext2 = "oe_mkext234fs ext2 ${EXTRA_IMAGECMD}"
> >  IMAGE_CMD_ext3 = "oe_mkext234fs ext3 ${EXTRA_IMAGECMD}"
> > -IMAGE_CMD_ext4 = "oe_mkext234fs ext4 ${EXTRA_IMAGECMD}"
> > +IMAGE_CMD_ext4 = "oe_mkext234fs ext4 -T default ${EXTRA_IMAGECMD}"
> >
> >  MIN_BTRFS_SIZE ?= "16384"
> >  IMAGE_CMD_btrfs () {
> > --
> > 2.25.1
> >
>
> >
> > 
> >
>
>
> --
> Alexandre Belloni, co-owner and COO, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com