[U-Boot] [PATCH v2 00/99] ram: rk3399: Add LPDDR4 support

[Jagan Teki]

Hi Kever,

On Thu, Jul 4, 2019 at 3:57 PM Kever Yang <kever.yang at rock-chips.com> wrote:
>
> Jagan,
>
>
> On 06/26/2019 06:22 PM, Jagan Teki wrote:
>
> On Wed, Jun 26, 2019 at 12:12 AM Ezequiel Garcia
> <ezequiel at vanguardiasur.com.ar> wrote:
>
> Hi Jagan,
>
> Thanks for your hard work. I'm sure everyone in the Rockchip community
> is excited about finally having this support in U-Boot.
>
> On Tue, 25 Jun 2019 at 12:46, Jagan Teki <jagan at amarulasolutions.com> wrote:
> [..]
>
> Was it absolutely necessary to split these changes into 99 commits? I
> believe at least some of them can be squashed. Reviewing 99 patches
> isn't feasible.
>
> Squashed, I'm not sure because the patches were created to satisfy the
> bisectability and travis-ci, if you find any please feel to comment.
> About the commit count, I have mentioned in v1, the idea of having
> many commits in one series to have all lpddr4(-related) changes in one
> place and also all the commit has incremental approach of supporting
> rank detection and lpddr4. If require I'm open to sent next versions
> as multiple series, no problem on that.
>
> I strongly agree with Vasily, and I don't think multiple series makes it any
> better.
>
> What's the reason for having two commits for:
>
> "ram: rk3399: Set lpddr4 MR3" and "ram: rk3399: Set lpddr4 MR12" ?
>
> These are individual lpddr4 set rate registers to support, each one is
> independent on it' own initialization and more over on the whole, it
> is critical to review.
>
> Or splitting all the "ram: rk3399: Add ... macro" ?
>
> You mean the patches 13 to 20 same like above each one has it's own
> meaning. It is not meaningful to squash them all.
>
>
> 99 patches is really too much, but I'm not sure how smaller it can be.
> Reference to kernel document, it suggest not more than 15 at one time:

Agreed.

>     NO!!!! No more huge patch bombs to linux-kernel at vger.kernel.org people! <https://lkml.org/lkml/2005/7/11/336> But please note that I don't think split this into different series make any sense. But maybe you can try to squash as much as you can. eg. the update for dram_all_config may able to squash into one patch,
> and some new MACRO and its reference code may be able to squash.
> So it depends on how you define about _logical change_.
> I'm not sure if this have happen in the history of U-Boot mailing list, but
> I think this big patch set will be complained by many people if this is send to
> kernel.

I don't mean to split the lpddr4 changes into multiple series. what
I'm trying to say here is this series has patches that support code
warnings, rank detection. Since each of them has it own identical
features, I'm planning to send them first. and will squash what it
require on lpddr4.

Will that be okay for you?