[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <fb31a159-6d2e-6c9a-439f-f19ef4fd4732@canonical.com>
Date: Sun, 5 Dec 2021 17:24:18 +0100
From: Krzysztof Kozlowski <krzysztof.kozlowski@...onical.com>
To: Sam Protsenko <semen.protsenko@...aro.org>,
Rob Herring <robh+dt@...nel.org>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc: Jaewon Kim <jaewon02.kim@...sung.com>,
Chanho Park <chanho61.park@...sung.com>,
David Virag <virag.david003@...il.com>,
Youngmin Nam <youngmin.nam@...sung.com>,
devicetree@...r.kernel.org, linux-serial@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
linux-samsung-soc@...r.kernel.org
Subject: Re: [PATCH v3 0/5] soc: samsung: Add USI driver
On 04/12/2021 20:57, Sam Protsenko wrote:
> USIv2 IP-core provides selectable serial protocol (UART, SPI or
> High-Speed I2C); only one can be chosen at a time. This series
> implements USIv2 driver, which allows one to select particular USI
> function in device tree, and also performs USI block initialization.
>
> With that driver implemented, it's not needed to do USI initialization
> in protocol drivers anymore, so that code is removed from the serial
> driver.
>
> Because USI driver is tristate (can be built as a module), serial driver
> was reworked so it's possible to use its console part as a module too.
> This way we can load serial driver module from user space and still have
> serial console functional.
>
> Design features:
> - "reg" property contains USI registers start address (0xc0 offset);
> it's used in the driver to access USI_CON and USI_OPTION registers.
> This way all USI initialization (reset, HWACG, etc) can be done in
> USIv2 driver separately, rather than duplicating that code over
> UART/SPI/I2C drivers
> - System Register (system controller node) and its SW_CONF register
> offset are provided in "samsung,sysreg" property; it's used to
> select USI function (protocol to be used)
> - USI function is specified in "samsung,mode" property; integer value
> is used to simplify parsing
> - there is "samsung,clkreq-on" bool property, which makes driver
> disable HWACG control (needed for UART to work properly)
> - PCLK and IPCLK clocks are both provided to USI node; apparently both
> need to be enabled to access USI registers
> - protocol nodes are embedded (as a child nodes) in USI node; it
> allows correct init order, and reflects HW properly
> - USI driver is a tristate: can be also useful from Android GKI
> requirements point of view
> - driver functions are implemented with further development in mind:
> - we might want to add some DebugFs interface later
> - some functions might need to be revealed to serial drivers with
> EXPORT_SYMBOL(), and provide somehow pointer to needed USI driver
> instance
> - another USI revisions could be added (like USIv1)
>
> Changes in v3:
> - Renamed compatible from samsung,exynos-usi-v2 to samsung,exynos850-usi
> - Used clk_bulk API instead of handling each clock separately
> - Spell check fixes and coding style fixes
> - Improved dt-bindings doc
>
> Changes in v2:
> - Renamed all 'usi_v2' wording to just 'usi' everywhere
> - Removed patches adding dependency on EXYNOS_USI for UART/I2C/SPI
> drivers
> - Added patch: "tty: serial: samsung: Fix console registration from
> module"
> - Combined dt-bindings doc and dt-bindings header patches
> - Reworked USI driver to be ready for USIv1 addition
> - Improved dt-bindings
> - Added USI_V2_NONE mode value
>
> Sam Protsenko (5):
> dt-bindings: soc: samsung: Add Exynos USI bindings
> soc: samsung: Add USI driver
> tty: serial: samsung: Remove USI initialization
> tty: serial: samsung: Enable console as module
> tty: serial: samsung: Fix console registration from module
>
All this looks good to me. The serial driver changes should come
together with this one (usi driver is now a dependency for them). If I
am correct, mention this please in future cover letter (if there is such).
I will still need DTSI changes for Exynos Auto v9 and confirmation that
is not being used downstream and breaking DTB ABI is okay. Because this
will be a non-bisctable and also a DTB ABI break.
Best regards,
Krzysztof
Powered by blists - more mailing lists