[PATCH 0/5] ARM: orion5x/dove/mv78xx0 multiplatform
dv at vollmann.ch
Fri Dec 11 15:18:43 PST 2015
On 12/11/15 22:33, Arnd Bergmann wrote:
> On Thursday 10 December 2015 14:39:27 Detlef Vollmann wrote:
>> # Automatically generated file; DO NOT EDIT.
>> # Linux/arm 4.4.0-rc3 Kernel Configuration
> Just for fun, a few things you can probably disable without
> much downsides:
> SYSVIPC could be disabled in principle for most embedded systems
I was a big fan of SYSV IPC and still sometimes prefer SYSV shm
over mmap. But yes, for this system I probably don't need it.
(Though I have a dim memory that one of the userspace programs
required it, but I will try without again.)
I need swap for the case I have hard disks mounted and run
checks on them.
>> # CONFIG_RELAY is not set
> SYSFS_DEPRECATED can probably go,
> initramfs would be smaller than initrd
But initramfs doesn't accept a squashfs.
The trick here is that the same squashfs is first used as initrd
and then as rootfs directly from flash (and the RAM of the initrd is
>> # CONFIG_SLUB is not set
>> # CONFIG_SLOB is not set
>> # CONFIG_SYSTEM_DATA_VERIFICATION is not set
>> # CONFIG_PROFILING is not set
> SLOB would be smaller than SLAB,
I didn't know this.
> kexec is often not needed.
This is the way I test new kernels.
I don't have JTAG mounted on this device and therefore don't
want to risk to replace the U-Boot. Unfortunately the installed
U-Boot doesn't support the network and uploading kernel and initrd
through the serial port isn't very reliable...
> I have no idea why SWIOTLB is always enabled on ARM, I suspect that's
> a bug, and disabling it would save some space (need to change arm/Kconfig)
Ok, I'll test it.
>> # User Modules And Translation Layers
> I suspect you don't use MTD_BLOCK
I believe SQUASHFS as well as MINIX FS require it.
>> # CONFIG_BLK_DEV_NULL_BLK is not set
>> # CONFIG_BLK_DEV_PCIESSD_MTIP32XX is not set
>> # CONFIG_BLK_CPQ_CISS_DA is not set
>> # CONFIG_BLK_DEV_DAC960 is not set
>> # CONFIG_BLK_DEV_UMEM is not set
>> # CONFIG_BLK_DEV_COW_COMMON is not set
>> # CONFIG_BLK_DEV_CRYPTOLOOP is not set
>> # CONFIG_BLK_DEV_DRBD is not set
>> # CONFIG_BLK_DEV_NBD is not set
>> # CONFIG_BLK_DEV_SX8 is not set
> Loop and RAM could be loadable modules, if that saves space in the
> end. If modules are actually larger, you can have everything built-in.
The flash partitioning is fix on this system (without my own U-Boot).
So I pushed as much as possible into the kernel binary to save
the other partition for real userspace programs. I didn't succeed
(XFS and btrfs are too big) but still try to cram as much as possible
into the kernel binary. My current rootfs spare is less than 100K
and I need to care for updates...
>> # Serial drivers
>> # CONFIG_SERIAL_8250_DEPRECATED_OPTIONS is not set
>> # CONFIG_SERIAL_8250_PCI is not set
>> # CONFIG_SERIAL_8250_EXTENDED is not set
> No need for 8250_FSL or 8250_DMA on your hardware
Thanks (I don't have proper documentation for this CPU).
>> # Memory mapped GPIO drivers
>> # CONFIG_GPIO_74XX_MMIO is not set
>> # CONFIG_GPIO_ALTERA is not set
>> # CONFIG_GPIO_DWAPB is not set
>> # CONFIG_GPIO_EM is not set
>> # CONFIG_GPIO_GENERIC_PLATFORM is not set
>> # CONFIG_GPIO_GRGPIO is not set
>> # CONFIG_GPIO_VX855 is not set
>> # CONFIG_GPIO_XILINX is not set
>> # CONFIG_GPIO_ZEVIO is not set
>> # CONFIG_GPIO_ZX is not set
> It looks like you get two GPIO drivers for the same hardware.
> We should probably change this so you only need one of the two.
Yes. I was surprised when I saw GPIO_MVEBU but ARCH_ORION5X turns
it always on.
>> # USB Host Controller Drivers
>> # CONFIG_USB_C67X00_HCD is not set
>> # CONFIG_USB_XHCI_HCD is not set
>> # CONFIG_USB_EHCI_TT_NEWSCHED is not set
> You probably don't have a PCI EHCI
I want to be able to run the rootfs from USB as a fallback.
>> # PHY Subsystem
> probably not needed since you have no specific driver for it.
>> # File systems
> EXT2 no longer needs to be built when EXT4 is enabled, you can just
> use the EXT4 code to mount EXT2 file systems.
Right now I have a bug with EXT4 (I can't delete a 200GB file).
So I probably trust the EXT2 code more (but I still want to be
able to have an EXT4 FS).
>> # CONFIG_XFS_QUOTA is not set
> XFS is huge. If you don't need it, that would likely save more than anything
I know. And btrfs is even worse (most features can't be configured off).
But this is a backup station and I want the options of the different
> minixfs is almost never needed, but small.
There are two flash partitions with the persistent system parameters.
These use MINIX FS (not my design).
>> # CONFIG_OMFS_FS is not set
>> # CONFIG_HPFS_FS is not set
>> # CONFIG_QNX4FS_FS is not set
>> # CONFIG_QNX6FS_FS is not set
>> # CONFIG_ROMFS_FS is not set
>> # CONFIG_PSTORE is not set
>> # CONFIG_SYSV_FS is not set
>> # CONFIG_UFS_FS is not set
> NFS and NFSD are also pretty big, but I assume you need them.
Thanks for your effort, but some time ago I hoped to get rid of
all modules (to save some space for userspace) and went through
the whole configuration. So there are at least no big chunks
to be removed...
More information about the linux-arm-kernel