diff mbox series

scripts: gen_image_generic: allow FAT fs on kernel partition for non-GPT targets

Message ID 20240328170108.13758-1-tmn505@terefe.re
State Superseded
Headers show
Series scripts: gen_image_generic: allow FAT fs on kernel partition for non-GPT targets | expand

Commit Message

Tomasz Maciej Nowak March 28, 2024, 5 p.m. UTC
From: Tomasz Maciej Nowak <tmn505@gmail.com>

Some old or proprietary bootloader recognize only FAT file system
variants on storage devices with MBR, unfortunately script ties
format of kernel partition to GPT partition table. So, allow kernel
partition file system to be FAT16 or FAT32 if appropriate type is set
in partition table.

Signed-off-by: Tomasz Maciej Nowak <tmn505@gmail.com>
---
 scripts/gen_image_generic.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Paul D March 29, 2024, 3:32 p.m. UTC | #1
Recommend avoiding -a and -o params.

Use instead e.g.

[ -n "$GUID" ] || [ "$KERNELPARTTYPE" = "6" ] || [ "$KERNELPARTTYPE" = "c" ]

https://www.shellcheck.net/wiki/SC2166



On 2024-03-28 18:00, Tomasz Maciej Nowak wrote:
> From: Tomasz Maciej Nowak <tmn505@gmail.com>
> 
> Some old or proprietary bootloader recognize only FAT file system
> variants on storage devices with MBR, unfortunately script ties
> format of kernel partition to GPT partition table. So, allow kernel
> partition file system to be FAT16 or FAT32 if appropriate type is set
> in partition table.
> 
> Signed-off-by: Tomasz Maciej Nowak <tmn505@gmail.com>
> ---
>   scripts/gen_image_generic.sh | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/scripts/gen_image_generic.sh b/scripts/gen_image_generic.sh
> index 11e40f38868f..44837fde1e12 100755
> --- a/scripts/gen_image_generic.sh
> +++ b/scripts/gen_image_generic.sh
> @@ -49,7 +49,7 @@ dos_dircopy() {
>   [ -n "$PADDING" ] && dd if=/dev/zero of="$OUTPUT" bs=512 seek="$ROOTFSOFFSET" conv=notrunc count="$ROOTFSSIZE"
>   dd if="$ROOTFSIMAGE" of="$OUTPUT" bs=512 seek="$ROOTFSOFFSET" conv=notrunc
>   
> -if [ -n "$GUID" ]; then
> +if [ -n "$GUID" -o "$KERNELPARTTYPE" = "6" -o "$KERNELPARTTYPE" = "c" ]; then
>       [ -n "$PADDING" ] && dd if=/dev/zero of="$OUTPUT" bs=512 seek="$((ROOTFSOFFSET + ROOTFSSIZE))" conv=notrunc count="$sect"
>       mkfs.fat --invariant -n kernel -C "$OUTPUT.kernel" -S 512 "$((KERNELSIZE / 1024))"
>       LC_ALL=C dos_dircopy "$KERNELDIR" /
Elliott Mitchell March 29, 2024, 6:08 p.m. UTC | #2
On Fri, Mar 29, 2024 at 04:32:18PM +0100, Paul D wrote:
> Recommend avoiding -a and -o params.
> 
> Use instead e.g.
> 
> [ -n "$GUID" ] || [ "$KERNELPARTTYPE" = "6" ] || [ "$KERNELPARTTYPE" = "c" ]
> 
> https://www.shellcheck.net/wiki/SC2166

The examples pointed to amounted to "be careful with untrusted input to
shell scripts".  Build systems must already be pretty much 100% trusted.
If someone slips something problematic into one there is pretty much
nothing to be done about it.


I've been getting the feeling the whole community is slowly trying to
recreate SunOS 5.7^WSolaris 2.7^WSoliaris 7 (used to be a Sun Thiing,
but now everyone's version numbers tend to be inflated).  Where
/bin/false, /bin/true and /bin/test were all Korne Shell scripts.

This nominally saved development work since Korne shell has
implementations of all these internally.  Problem is this killed
performance for any shell script /not/ written in a shell with those as
built-in.  While Korne shell is very fast once it has finished parsing
its input, it is slow at parsing its scripts.

Having `gunzip` be a shell script seems headed in this direction.  The
above seems a similar sort of situation, adding an extra fork()/execve()
to avoid warnings.

I'm placing SC2166 on my disregard list.  Yes, this is nominally not well
defined, but unlike most warnings this one has a major impact on
performance.

(fork()/execve() are the two most expensive routinely used system calls)
Tomasz Maciej Nowak April 3, 2024, 4:30 p.m. UTC | #3
W dniu 29.03.2024 o 19:08, Elliott Mitchell pisze:
> On Fri, Mar 29, 2024 at 04:32:18PM +0100, Paul D wrote:
>> Recommend avoiding -a and -o params.
>>
>> Use instead e.g.
>>
>> [ -n "$GUID" ] || [ "$KERNELPARTTYPE" = "6" ] || [ "$KERNELPARTTYPE" = "c" ]
>>
>> https://www.shellcheck.net/wiki/SC2166
> 
> The examples pointed to amounted to "be careful with untrusted input to
> shell scripts".  Build systems must already be pretty much 100% trusted.
> If someone slips something problematic into one there is pretty much
> nothing to be done about it.
> 
> 
> I've been getting the feeling the whole community is slowly trying to
> recreate SunOS 5.7^WSolaris 2.7^WSoliaris 7 (used to be a Sun Thiing,
> but now everyone's version numbers tend to be inflated).  Where
> /bin/false, /bin/true and /bin/test were all Korne Shell scripts.
> 
> This nominally saved development work since Korne shell has
> implementations of all these internally.  Problem is this killed
> performance for any shell script /not/ written in a shell with those as
> built-in.  While Korne shell is very fast once it has finished parsing
> its input, it is slow at parsing its scripts.
> 
> Having `gunzip` be a shell script seems headed in this direction.  The
> above seems a similar sort of situation, adding an extra fork()/execve()
> to avoid warnings.
> 
> I'm placing SC2166 on my disregard list.  Yes, this is nominally not well
> defined, but unlike most warnings this one has a major impact on
> performance.
> 
> (fork()/execve() are the two most expensive routinely used system calls)
> 

Thanks, I did consider the execution penalty but as this script is small I decided
to go with Paul's suggestion, and as bonus, just to avoid future noise around
this, when someone else validates this script with shellcheck.
diff mbox series

Patch

diff --git a/scripts/gen_image_generic.sh b/scripts/gen_image_generic.sh
index 11e40f38868f..44837fde1e12 100755
--- a/scripts/gen_image_generic.sh
+++ b/scripts/gen_image_generic.sh
@@ -49,7 +49,7 @@  dos_dircopy() {
 [ -n "$PADDING" ] && dd if=/dev/zero of="$OUTPUT" bs=512 seek="$ROOTFSOFFSET" conv=notrunc count="$ROOTFSSIZE"
 dd if="$ROOTFSIMAGE" of="$OUTPUT" bs=512 seek="$ROOTFSOFFSET" conv=notrunc
 
-if [ -n "$GUID" ]; then
+if [ -n "$GUID" -o "$KERNELPARTTYPE" = "6" -o "$KERNELPARTTYPE" = "c" ]; then
     [ -n "$PADDING" ] && dd if=/dev/zero of="$OUTPUT" bs=512 seek="$((ROOTFSOFFSET + ROOTFSSIZE))" conv=notrunc count="$sect"
     mkfs.fat --invariant -n kernel -C "$OUTPUT.kernel" -S 512 "$((KERNELSIZE / 1024))"
     LC_ALL=C dos_dircopy "$KERNELDIR" /