Re: [PATCH v4 0/3] Add basic support for LS7A bridge chip

From: Tiezhu Yang
Date: Mon Apr 27 2020 - 01:42:29 EST


On 04/27/2020 12:57 PM, Jiaxun Yang wrote:
On Mon, 27 Apr 2020 09:31:54 +0800
Tiezhu Yang <yangtiezhu@xxxxxxxxxxx> wrote:

On 04/03/2020 05:29 PM, Tiezhu Yang wrote:
The LS7A bridge chip has been released for several years since the
second half of 2017, but it is not supported by the Linux mainline
kernel while it only works well with the Loongson internal kernel
version. When I update the latest version of Linux mainline kernel
on the Loongson 3A3000 CPU and LS7A bridge chip system, the boot
process failed and I feel depressed.

The LS7A bridge chip is used a lot with 3A3000 or 3A4000 CPU in
the most Loongson desktop and sever products, it is important to
support LS7A bridge chip by the Linux mainline kernel.

This patch series adds the basic support for the LS7A bridge chip,
the patch about vendor ID and SATA has been merged into the mainline
tree, the next work is to refactor the code about the interrupt
controller, and then power management and some other controller
device drivers.

By the way, if you want the boot process is successful (just for
test) on the Loongson 3A3000 CPU and LS7A bridge chip system,
you should not only apply these patches, but also need the support
for SATA and interrupt controller in the v1 patch series.

This patch series is based on mips-next.

If you have any questions and suggestions, please let me know.

Thanks,

Tiezhu Yang

v2:
- The split patch series about Loongson vendor ID and SATA
controller has been merged into the linux-block.git by Jens Axboe
[1].

- Think about using hierarchy IRQ domain in the patch of
interrupt controller, and this maybe depend on the patch series by
Jiaxun ("Modernize Loongson64 Machine"), so the patch about
interrupt is not included in this v2 patch series.

v3:
- The split patch series about Loongson vendor ID and SATA
controller has been merged into the mainline tree [2]

- Modify the macro definition and add comment to make it easy to
read

- Move ls7a1000_pci_class_quirk() to fixup-loongson3.c

- Use PCI_VENDOR_ID_LOONGSON in pci_ids.h instead of 0x0014

v4:
- Use LS7A instead of Loongson 7A1000 in the description
- Use LS7A or ls7a instead of LS7A1000 or ls7a1000 in the code

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git/commit/?h=for-next&id=9acb9fe18d86
https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git/commit/?h=for-next&id=e49bd683e00b
[2]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=9acb9fe18d86
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e49bd683e00b

Tiezhu Yang (3):
MIPS: Loongson: Get host bridge information
MIPS: Loongson: Add DMA support for LS7A
MIPS: Loongson: Add PCI support for LS7A

arch/mips/include/asm/mach-loongson64/boot_param.h | 20 +++++++
arch/mips/loongson64/dma.c | 9 ++--
arch/mips/loongson64/env.c | 22 ++++++++
arch/mips/loongson64/init.c | 17 ++++++
arch/mips/pci/fixup-loongson3.c | 12 +++++
arch/mips/pci/ops-loongson3.c | 63
++++++++++++++++++++-- 6 files changed, 136 insertions(+), 7
deletions(-)
Hi Thomas,

Could you please apply the following two patches to mips-next?

[v4,1/3] MIPS: Loongson: Get host bridge information
https://lore.kernel.org/patchwork/patch/1220009/

[v4,2/3] MIPS: Loongson: Add DMA support for LS7A
https://lore.kernel.org/patchwork/patch/1220010/
Sorry but I really don't like the DMA patch, hard coding a config
register in platform code is not necessarily a good idea, it
create painful hell for adding new platform support.

I'm trying very hard to let all devices go through DeviceTree.

I'd suggest you to limit DMA capability on LS7A PCH instead of doing
the hack.

Hi Jiaxun,

Thanks for your suggestion, let me rethink how to properly support DMA.

Do you think the following patch could be applied first?

[v4,1/3] MIPS: Loongson: Get host bridge information
https://lore.kernel.org/patchwork/patch/1220009/

Thanks,
Tiezhu Yang


Or if you think the function is necessary, you can create a DeviceTree
node called "loongson,ls7a-syscon", and do whatever you like in
init_calls by parsing this node.

Also that will also block my upcoming LS2K support patches.
You can check my tree here[1].

Thanks.

[1]: https://github.com/FlyGoat/linux/commits/next-testing-2k

Thanks,
Tiezhu Yang

--
Jiaxun Yang