Re: [PATCH v3 13/13] ARM: imxrt_defconfig: add i.MXRT family defconfig

[Jesse Taube]

On 11/25/21 16:42, Arnd Bergmann wrote:
> On Thu, Nov 25, 2021 at 10:14 PM Jesse Taube <mr.bossman075@xxxxxxxxx> wrote:
> > From: Giulio Benetti <giulio.benetti@xxxxxxxxxxxxxxxxxxxxxx>
> >
> > Add generic i.MXRT family defconfig.
> >
> > Signed-off-by: Giulio Benetti <giulio.benetti@xxxxxxxxxxxxxxxxxxxxxx>
> > Signed-off-by: Jesse Taube <Mr.Bossman075@xxxxxxxxx>
>
> I see a lot of things in here that probably should not be part of the kernel,
> either because they are rather obscure, or they take valuable memory:
>
> > ---
> >   arch/arm/configs/imxrt_defconfig | 157 +++++++++++++++++++++++++++++++
> >   1 file changed, 157 insertions(+)
> >   create mode 100644 arch/arm/configs/imxrt_defconfig
> >
> > diff --git a/arch/arm/configs/imxrt_defconfig b/arch/arm/configs/imxrt_defconfig
> > new file mode 100644
> > index 000000000000..d673745a5462
> > --- /dev/null
> > +++ b/arch/arm/configs/imxrt_defconfig
> > @@ -0,0 +1,157 @@
> > +# CONFIG_LOCALVERSION_AUTO is not set
> > +CONFIG_SYSVIPC=y
> > +CONFIG_USELIB=y
>
> You almost certainly won't want USELIB, and SYSVIPC support
> would only be useful for certain applications that you probably
> won't run.
>
> > +CONFIG_BSD_PROCESS_ACCT=y
> > +CONFIG_BSD_PROCESS_ACCT_V3=y
> > +CONFIG_PSI=y
> > +CONFIG_IKCONFIG=y
> > +CONFIG_IKCONFIG_PROC=y
>
> Probably nonen of these are helpful here.
>
> > +CONFIG_MEMCG=y
> > +CONFIG_BLK_CGROUP=y
> > +CONFIG_CFS_BANDWIDTH=y
> > +CONFIG_CGROUP_PIDS=y
> > +CONFIG_CGROUP_RDMA=y
> > +CONFIG_CGROUP_FREEZER=y
> > +CONFIG_CGROUP_DEVICE=y
> > +CONFIG_CGROUP_CPUACCT=y
> > +CONFIG_CGROUP_PERF=y
> > +CONFIG_CGROUP_BPF=y
> > +CONFIG_NAMESPACES=y
> > +CONFIG_USER_NS=y
> > +CONFIG_CHECKPOINT_RESTORE=y
>
> Same for control groups overall as well as checkpoint_restore
>
> > +CONFIG_RELAY=y
>
> There are a few drivers using CONFIG_RELAY, but I don't see you enable
> any of them,
> so this is not actually used.
this makes scene will change
> > +CONFIG_EXPERT=y
>
> Are you losing anything without EXPERT? If not, remove that
>
> > +CONFIG_SGETMASK_SYSCALL=y
> > +# CONFIG_FUTEX is not set
>
> Futex is probably one of the things you /do/ want.
There is a weird bug in futexs that reads from null_ptr, otherwise if 
you hard code that access isn't allowed it works. This seems to be a 
problem on other no-MMU platforms. I changed it to not have futexs like 
the stm32s, CONFIG_EXPERT is required for this option. If you want I can 
explain more about this but, i was going to do that as a separate patch.

> > +CONFIG_KALLSYMS_ALL=y
> > +CONFIG_PC104=y
>
> Turning off KALLSYMS_ALL may save a noticeable amount of RAM.
Intresting
> PC104 isn't actually that big, but it seems unlikely that you have
> that hardware.
>
> > +CONFIG_PARAVIRT=y
>
> You don't seem to enable XEN, so I don't think PARAVIRT is useful by itself.
>
> > +# CONFIG_ATAGS is not set
> > +CONFIG_CMDLINE="console=ttyS0 root=/dev/mmcblk0p2 rw earlycon rootwait"
>
> The command line should come from the boot loader, users probably have
> a different root device.
Presumably this will never be used but I have it there as an example for 
others, I can remove it if necessary.
> > +CONFIG_BLK_DEV_BSGLIB=y
> > +CONFIG_BLK_DEV_INTEGRITY=y
> > +CONFIG_BLK_DEV_ZONED=y
> > +CONFIG_BLK_DEV_THROTTLING=y
> > +CONFIG_BLK_WBT=y
> > +CONFIG_BLK_SED_OPAL=y
> > +CONFIG_PARTITION_ADVANCED=y
> > +CONFIG_BSD_DISKLABEL=y
> > +CONFIG_MINIX_SUBPARTITION=y
> > +CONFIG_SOLARIS_X86_PARTITION=y
> > +CONFIG_UNIXWARE_DISKLABEL=y
> > +CONFIG_LDM_PARTITION=y
>
> I don't see you using OPAL or any of the 1990's partition formats.
> ot set
>
> > +CONFIG_BINFMT_FLAT=y
>
> For the defconfig, you should probably have ELF_FDPIC enabled,
> not just FLAT.
>
> > +CONFIG_CLEANCACHE=y
> > +CONFIG_ZPOOL=y
> > +CONFIG_ZBUD=y
> > +CONFIG_Z3FOLD=y
>
> Do these work as expected on NOMMU?
>
> > +CONFIG_BLK_DEV_LOOP=y
> > +CONFIG_BLK_DEV_RAM=y
> > +CONFIG_BLK_DEV_RAM_COUNT=1
> > +CONFIG_BLK_DEV_RAM_SIZE=65536
>
> I don't think you can have a ramdisk larger than RAM ;-)
Oh that's funny I left that from testing something I will remove it.
> > +CONFIG_MEMORY=y
>
> No need to enable the subsystem when you don't enable any
> of its drivers.
>
> > +CONFIG_EXT2_FS=y
> > +CONFIG_EXT2_FS_XATTR=y
> > +CONFIG_EXT2_FS_POSIX_ACL=y
> > +CONFIG_EXT2_FS_SECURITY=y
> > +CONFIG_EXT3_FS=y
> > +CONFIG_EXT3_FS_POSIX_ACL=y
> > +CONFIG_EXT3_FS_SECURITY=y
>
> Never use EXT3 on eMMC, just use EXT4 instead to make
> the device actually live longer. You probably don't need to enable
> EXT2 support separately.
>
> > +# CONFIG_FILE_LOCKING is not set
> > +# CONFIG_DNOTIFY is not set
> > +CONFIG_QUOTA=y
>
> dnotify and locking seem more useful than quota here
>
> > +# CONFIG_PRINT_QUOTA_WARNING is not set
> > +CONFIG_AUTOFS4_FS=y
>
> > +CONFIG_CONFIGFS_FS=y
>
> I don't see anything using configfs
>
> > +CONFIG_LSM="yama,loadpin,integrity,apparmor"
>
> None of these are actually enabled as far as I can tell.
>
> > +CONFIG_DEBUG_INFO=y
> > +CONFIG_DEBUG_INFO_DWARF4=y
>
> If you use DWARF4, you probably want DEBUG_INFO_SPLIT as well,
> to reduce the vmlinux size.
>
>         Arnd
Thx for telling me about this i will try to use only the necessary ones.