All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jeffrey Hugo <jeffrey.l.hugo@gmail.com>
To: michael.srba@seznam.cz
Cc: Andy Gross <agross@kernel.org>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Rob Herring <robh+dt@kernel.org>, Stephen Boyd <sboyd@kernel.org>,
	Philipp Zabel <p.zabel@pengutronix.de>,
	Linus Walleij <linus.walleij@linaro.org>,
	Florian Fainelli <f.fainelli@gmail.com>,
	Arnd Bergmann <arnd@arndb.de>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Saravana Kannan <saravanak@google.com>,
	MSM <linux-arm-msm@vger.kernel.org>,
	linux-clk@vger.kernel.org, DTML <devicetree@vger.kernel.org>
Subject: Re: [PATCH v8 4/5] drivers: bus: add driver for initializing the SSC bus on (some) qcom SoCs
Date: Sun, 27 Feb 2022 13:25:14 -0700	[thread overview]
Message-ID: <CAOCk7NpXa03fzRSXEEwkBANZNXyb_LLAje1EMsaprUFv0ew2HQ@mail.gmail.com> (raw)
In-Reply-To: <20220220212034.9152-4-michael.srba@seznam.cz>

On Sun, Feb 20, 2022 at 11:12 PM <michael.srba@seznam.cz> wrote:
>
> From: Michael Srba <Michael.Srba@seznam.cz>
>
> Add bindings for the AHB bus which exposes the SCC block in the global
> address space. This bus (and the SSC block itself) is present on certain
> qcom SoCs.

"SCC" or "SSC"?

> In typical configuration, this bus (as some of the clocks and registers
> that we need to manipulate) is not accessible to Linux, and the resources
> on this bus are indirectly accessed by communicating with a hexagon CPU
> core residing in the SSC block. In this configuration, the hypervisor is
> the one performing the bus initialization for the purposes of bringing
> the haxagon CPU core out of reset.

"hexagon"

> However, it is possible to change the configuration, in which case this
> driver will initialize the bus.
>
> In combination with drivers for resources on the SSC bus, this driver can
> aid in debugging, and for example with a TLMM driver can be used to
> directly access SSC-dedicated GPIO pins, removing the need to commit
> to a particular usecase during hw design.
>
> Finally, until open firmware for the hexagon core is available, this
> approach allows for using sensors hooked up to SSC-dedicated GPIO pins
> on mainline Linux simply by utilizing the existing in-tree drivers for
> these sensors.
>
> Signed-off-by: Michael Srba <Michael.Srba@seznam.cz>
> ---
>  CHANGES:
>  - v2: none
>  - v3: fix clang warning
>  - v4: address the issues pointed out in the review
>  - v5: none
>  - v6: restore alphabetic ordering in Makefile against v5.17-rc4
>  - v7: use imperative in commit message
>  - v8: none
> ---
>  drivers/bus/Kconfig              |   6 +
>  drivers/bus/Makefile             |   1 +
>  drivers/bus/qcom-ssc-block-bus.c | 383 +++++++++++++++++++++++++++++++
>  3 files changed, 390 insertions(+)
>  create mode 100644 drivers/bus/qcom-ssc-block-bus.c
>
> diff --git a/drivers/bus/Kconfig b/drivers/bus/Kconfig
> index 3c68e174a113..9e29d1da9a61 100644
> --- a/drivers/bus/Kconfig
> +++ b/drivers/bus/Kconfig
> @@ -152,6 +152,12 @@ config QCOM_EBI2
>           Interface 2, which can be used to connect things like NAND Flash,
>           SRAM, ethernet adapters, FPGAs and LCD displays.
>
> +config QCOM_SSC_BLOCK_BUS
> +       bool "Qualcomm SSC Block Bus Init Driver"
> +         help
> +         Say y here to enable support for initializing the bus that connects the SSC block's internal
> +         bus to the cNoC on (some) qcom SoCs

Please define "SSC" and "cNoC" for those not familiar with the terms.
If someone has questions about this config item or driver, this is
probably the first place they are going to look.

Also, a depends on ARCH_QCOM will probably fix the s390 build issue.

> +
>  config SUN50I_DE2_BUS
>         bool "Allwinner A64 DE2 Bus Driver"
>           default ARM64
> diff --git a/drivers/bus/Makefile b/drivers/bus/Makefile
> index 52c2f35a26a9..e6756e83a9c4 100644
> --- a/drivers/bus/Makefile
> +++ b/drivers/bus/Makefile
> @@ -25,6 +25,7 @@ obj-$(CONFIG_OMAP_INTERCONNECT)       += omap_l3_smx.o omap_l3_noc.o
>
>  obj-$(CONFIG_OMAP_OCP2SCP)     += omap-ocp2scp.o
>  obj-$(CONFIG_QCOM_EBI2)                += qcom-ebi2.o
> +obj-$(CONFIG_QCOM_SSC_BLOCK_BUS)       += qcom-ssc-block-bus.o
>  obj-$(CONFIG_SUN50I_DE2_BUS)   += sun50i-de2.o
>  obj-$(CONFIG_SUNXI_RSB)                += sunxi-rsb.o
>  obj-$(CONFIG_OF)               += simple-pm-bus.o
> diff --git a/drivers/bus/qcom-ssc-block-bus.c b/drivers/bus/qcom-ssc-block-bus.c
> new file mode 100644
> index 000000000000..e489f5614e90
> --- /dev/null
> +++ b/drivers/bus/qcom-ssc-block-bus.c
> @@ -0,0 +1,383 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +// Copyright (c) 2021, Michael Srba
> +
> +#include <linux/clk.h>
> +#include <linux/delay.h>
> +#include <linux/io.h>
> +#include <linux/mfd/syscon.h>
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_clock.h>
> +#include <linux/pm_domain.h>
> +#include <linux/pm_runtime.h>
> +#include <linux/regmap.h>
> +#include <linux/reset.h>
> +
> +/* AXI Halt Register Offsets */
> +#define AXI_HALTREQ_REG                        0x0
> +#define AXI_HALTACK_REG                        0x4
> +#define AXI_IDLE_REG                   0x8
> +
> +static const char *const qcom_ssc_block_pd_names[] = {
> +       "ssc_cx",
> +       "ssc_mx"
> +};
> +
> +struct qcom_ssc_block_bus_data {
> +       int num_pds;
> +       const char *const *pd_names;
> +       struct device *pds[ARRAY_SIZE(qcom_ssc_block_pd_names)];
> +       char __iomem *reg_mpm_sscaon_config0; // MPM - msm power manager; AON - always-on
> +       char __iomem *reg_mpm_sscaon_config1; // that's as much as we know about these

I'm only going to comment on this once - C++ style comments in code
are against the Linux coding convention.

> +       struct regmap *halt_map;
> +       u32 ssc_axi_halt;
> +       struct clk *xo_clk;
> +       struct clk *aggre2_clk;
> +       struct clk *gcc_im_sleep_clk;
> +       struct clk *aggre2_north_clk;
> +       struct clk *ssc_xo_clk;
> +       struct clk *ssc_ahbs_clk;
> +       struct reset_control *ssc_bcr;
> +       struct reset_control *ssc_reset;

I'm curious, have you run pahole on this struct?  I suspect there is a
lot of useless padding in it.

> +};
> +
> +static void reg32_set_bits(char __iomem *reg, u32 value)
> +{
> +       u32 tmp = ioread32(reg);
> +
> +       iowrite32(tmp | value, reg);
> +}
> +
> +static void reg32_clear_bits(char __iomem *reg, u32 value)
> +{
> +       u32 tmp = ioread32(reg);
> +
> +       iowrite32(tmp & (~value), reg);
> +}
> +
> +
> +static int qcom_ssc_block_bus_init(struct device *dev)
> +{
> +       int ret;
> +
> +       struct qcom_ssc_block_bus_data *data = dev_get_drvdata(dev);
> +
> +       ret = clk_prepare_enable(data->xo_clk);
> +       if (ret) {
> +               dev_err(dev, "error enabling xo_clk: %d\n", ret);
> +               goto err_xo_clk;
> +       }
> +
> +       ret = clk_prepare_enable(data->aggre2_clk);
> +       if (ret) {
> +               dev_err(dev, "error enabling aggre2_clk: %d\n", ret);
> +               goto err_aggre2_clk;
> +       }
> +
> +       ret = clk_prepare_enable(data->gcc_im_sleep_clk);
> +       if (ret) {
> +               dev_err(dev, "error enabling gcc_im_sleep_clk: %d\n", ret);
> +               goto err_gcc_im_sleep_clk;
> +       }
> +
> +       reg32_clear_bits(data->reg_mpm_sscaon_config0, BIT(4) | BIT(5));
> +       reg32_clear_bits(data->reg_mpm_sscaon_config1, BIT(31));

This seems like magic.  I don't think you need to create #DEFINEs for
this, but maybe a comment about what bits your are clearing, and why?

> +
> +       ret = clk_prepare_enable(data->aggre2_north_clk);
> +       if (ret) {
> +               dev_err(dev, "error enabling aggre2_north_clk: %d\n", ret);
> +               goto err_aggre2_north_clk;
> +       }
> +
> +       ret = reset_control_deassert(data->ssc_reset);
> +       if (ret) {
> +               dev_err(dev, "error deasserting ssc_reset: %d\n", ret);
> +               goto err_ssc_reset;
> +       }
> +
> +       ret = reset_control_deassert(data->ssc_bcr);
> +       if (ret) {
> +               dev_err(dev, "error deasserting ssc_bcr: %d\n", ret);
> +               goto err_ssc_bcr;
> +       }
> +
> +       regmap_write(data->halt_map, data->ssc_axi_halt + AXI_HALTREQ_REG, 0);
> +
> +       ret = clk_prepare_enable(data->ssc_xo_clk);
> +       if (ret) {
> +               dev_err(dev, "error deasserting ssc_xo_clk: %d\n", ret);
> +               goto err_ssc_xo_clk;
> +       }
> +
> +       ret = clk_prepare_enable(data->ssc_ahbs_clk);
> +       if (ret) {
> +               dev_err(dev, "error deasserting ssc_ahbs_clk: %d\n", ret);
> +               goto err_ssc_ahbs_clk;
> +       }
> +
> +       return 0;
> +
> +err_ssc_ahbs_clk:
> +       clk_disable(data->ssc_xo_clk);
> +
> +err_ssc_xo_clk:
> +       regmap_write(data->halt_map, data->ssc_axi_halt + AXI_HALTREQ_REG, 1);
> +
> +       reset_control_assert(data->ssc_bcr);
> +
> +err_ssc_bcr:
> +       reset_control_assert(data->ssc_reset);
> +
> +err_ssc_reset:
> +       clk_disable(data->aggre2_north_clk);
> +
> +err_aggre2_north_clk:
> +       reg32_set_bits(data->reg_mpm_sscaon_config0, BIT(4) | BIT(5));
> +       reg32_set_bits(data->reg_mpm_sscaon_config1, BIT(31));
> +
> +       clk_disable(data->gcc_im_sleep_clk);
> +
> +err_gcc_im_sleep_clk:
> +       clk_disable(data->aggre2_clk);
> +
> +err_aggre2_clk:
> +       clk_disable(data->xo_clk);
> +
> +err_xo_clk:
> +       return ret;
> +}
> +
> +static void qcom_ssc_block_bus_deinit(struct device *dev)
> +{
> +       int ret;
> +
> +       struct qcom_ssc_block_bus_data *data = dev_get_drvdata(dev);
> +
> +       clk_disable(data->ssc_xo_clk);
> +       clk_disable(data->ssc_ahbs_clk);
> +
> +       ret = reset_control_assert(data->ssc_bcr);
> +       if (ret)
> +               dev_err(dev, "error asserting ssc_bcr: %d\n", ret);
> +
> +       regmap_write(data->halt_map, data->ssc_axi_halt + AXI_HALTREQ_REG, 1);
> +
> +       reg32_set_bits(data->reg_mpm_sscaon_config1, BIT(31));
> +       reg32_set_bits(data->reg_mpm_sscaon_config0, BIT(4) | BIT(5));
> +
> +       ret = reset_control_assert(data->ssc_reset);
> +       if (ret)
> +               dev_err(dev, "error asserting ssc_reset: %d\n", ret);
> +
> +       clk_disable(data->gcc_im_sleep_clk);
> +
> +       clk_disable(data->aggre2_north_clk);
> +
> +       clk_disable(data->aggre2_clk);
> +       clk_disable(data->xo_clk);
> +}
> +
> +
> +static int qcom_ssc_block_bus_pds_attach(struct device *dev, struct device **pds,
> +                                        const char *const *pd_names, size_t num_pds)
> +{
> +       int ret;
> +       int i;
> +
> +       for (i = 0; i < num_pds; i++) {
> +               pds[i] = dev_pm_domain_attach_by_name(dev, pd_names[i]);
> +               if (IS_ERR_OR_NULL(pds[i])) {
> +                       ret = PTR_ERR(pds[i]) ? : -ENODATA;
> +                       goto unroll_attach;
> +               }
> +       }
> +
> +       return num_pds;
> +
> +unroll_attach:
> +       for (i--; i >= 0; i--)
> +               dev_pm_domain_detach(pds[i], false);
> +
> +       return ret;
> +};
> +
> +static void qcom_ssc_block_bus_pds_detach(struct device *dev, struct device **pds, size_t num_pds)
> +{
> +       int i;
> +
> +       for (i = 0; i < num_pds; i++)
> +               dev_pm_domain_detach(pds[i], false);
> +}
> +
> +static int qcom_ssc_block_bus_pds_enable(struct device **pds, size_t num_pds)
> +{
> +       int ret;
> +       int i;
> +
> +       for (i = 0; i < num_pds; i++) {
> +               dev_pm_genpd_set_performance_state(pds[i], INT_MAX);
> +               ret = pm_runtime_get_sync(pds[i]);
> +               if (ret < 0)
> +                       goto unroll_pd_votes;
> +       }
> +
> +       return 0;
> +
> +unroll_pd_votes:
> +       for (i--; i >= 0; i--) {
> +               dev_pm_genpd_set_performance_state(pds[i], 0);
> +               pm_runtime_put(pds[i]);
> +       }
> +
> +       return ret;
> +};
> +
> +static void qcom_ssc_block_bus_pds_disable(struct device **pds, size_t num_pds)
> +{
> +       int i;
> +
> +       for (i = 0; i < num_pds; i++) {
> +               dev_pm_genpd_set_performance_state(pds[i], 0);
> +               pm_runtime_put(pds[i]);
> +       }
> +}
> +
> +static int qcom_ssc_block_bus_probe(struct platform_device *pdev)
> +{
> +       struct qcom_ssc_block_bus_data *data;
> +       struct device_node *np = pdev->dev.of_node;
> +       struct of_phandle_args halt_args;
> +       struct resource *res;
> +       int ret;
> +
> +       data = devm_kzalloc(&pdev->dev, sizeof(*data), GFP_KERNEL);
> +       if (!data)
> +               return -ENOMEM;
> +
> +       platform_set_drvdata(pdev, data);
> +
> +       data->pd_names = qcom_ssc_block_pd_names;
> +       data->num_pds = ARRAY_SIZE(qcom_ssc_block_pd_names);
> +
> +       // power domains
> +       ret = qcom_ssc_block_bus_pds_attach(&pdev->dev, data->pds, data->pd_names, data->num_pds);
> +       if (ret < 0)
> +               return dev_err_probe(&pdev->dev, ret, "error when attaching power domains\n");
> +
> +       ret = qcom_ssc_block_bus_pds_enable(data->pds, data->num_pds);
> +       if (ret < 0)
> +               return dev_err_probe(&pdev->dev, ret, "error when enabling power domains\n");
> +
> +       // the meaning of the bits in these two registers is sadly not documented,
> +       // the set/clear operations are just copying what qcom does
> +       res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "mpm_sscaon_config0");
> +       data->reg_mpm_sscaon_config0 = devm_ioremap_resource(&pdev->dev, res);
> +       if (IS_ERR(data->reg_mpm_sscaon_config0))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->reg_mpm_sscaon_config0),
> +                                    "Failed to ioremap mpm_sscaon_config0\n");
> +
> +       res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "mpm_sscaon_config1");
> +       data->reg_mpm_sscaon_config1 = devm_ioremap_resource(&pdev->dev, res);
> +       if (IS_ERR(data->reg_mpm_sscaon_config1))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->reg_mpm_sscaon_config1),
> +                                    "Failed to ioremap mpm_sscaon_config1\n");
> +
> +       // resets
> +       data->ssc_bcr = devm_reset_control_get_exclusive(&pdev->dev, "ssc_bcr");
> +       if (IS_ERR(data->ssc_bcr))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->ssc_bcr),
> +                                    "Failed to acquire reset: scc_bcr\n");
> +
> +       data->ssc_reset = devm_reset_control_get_exclusive(&pdev->dev, "ssc_reset");
> +       if (IS_ERR(data->ssc_reset))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->ssc_reset),
> +                                    "Failed to acquire reset: ssc_reset:\n");
> +
> +       // clocks
> +       data->xo_clk = devm_clk_get(&pdev->dev, "xo");
> +       if (IS_ERR(data->xo_clk))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->xo_clk),
> +                                    "Failed to get clock: xo\n");
> +
> +       data->aggre2_clk = devm_clk_get(&pdev->dev, "aggre2");
> +       if (IS_ERR(data->aggre2_clk))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->aggre2_clk),
> +                                    "Failed to get clock: aggre2\n");
> +
> +       data->gcc_im_sleep_clk = devm_clk_get(&pdev->dev, "gcc_im_sleep");
> +       if (IS_ERR(data->gcc_im_sleep_clk))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->gcc_im_sleep_clk),
> +                                    "Failed to get clock: gcc_im_sleep\n");
> +
> +       data->aggre2_north_clk = devm_clk_get(&pdev->dev, "aggre2_north");
> +       if (IS_ERR(data->aggre2_north_clk))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->aggre2_north_clk),
> +                                    "Failed to get clock: aggre2_north\n");
> +
> +       data->ssc_xo_clk = devm_clk_get(&pdev->dev, "ssc_xo");
> +       if (IS_ERR(data->ssc_xo_clk))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->ssc_xo_clk),
> +                                    "Failed to get clock: ssc_xo\n");
> +
> +       data->ssc_ahbs_clk = devm_clk_get(&pdev->dev, "ssc_ahbs");
> +       if (IS_ERR(data->ssc_ahbs_clk))
> +               return dev_err_probe(&pdev->dev, PTR_ERR(data->ssc_ahbs_clk),
> +                                    "Failed to get clock: ssc_ahbs\n");
> +
> +       ret = of_parse_phandle_with_fixed_args(pdev->dev.of_node, "qcom,halt-regs", 1, 0,
> +                                              &halt_args);
> +       if (ret < 0)
> +               return dev_err_probe(&pdev->dev, ret, "Failed to parse qcom,halt-regs\n");
> +
> +       data->halt_map = syscon_node_to_regmap(halt_args.np);
> +       of_node_put(halt_args.np);
> +       if (IS_ERR(data->halt_map))
> +               return PTR_ERR(data->halt_map);
> +
> +       data->ssc_axi_halt = halt_args.args[0];
> +
> +       qcom_ssc_block_bus_init(&pdev->dev);
> +
> +       of_platform_populate(np, NULL, NULL, &pdev->dev);
> +
> +       return 0;
> +}
> +
> +static int qcom_ssc_block_bus_remove(struct platform_device *pdev)
> +{
> +       struct qcom_ssc_block_bus_data *data = platform_get_drvdata(pdev);
> +
> +       qcom_ssc_block_bus_deinit(&pdev->dev);
> +
> +       iounmap(data->reg_mpm_sscaon_config0);
> +       iounmap(data->reg_mpm_sscaon_config1);
> +
> +       qcom_ssc_block_bus_pds_disable(data->pds, data->num_pds);
> +       qcom_ssc_block_bus_pds_detach(&pdev->dev, data->pds, data->num_pds);
> +       pm_runtime_disable(&pdev->dev);
> +       pm_clk_destroy(&pdev->dev);
> +
> +       return 0;
> +}
> +
> +static const struct of_device_id qcom_ssc_block_bus_of_match[] = {
> +       { .compatible = "qcom,ssc-block-bus", },
> +       { /* sentinel */ }
> +};
> +MODULE_DEVICE_TABLE(of, qcom_ssc_block_bus_of_match);
> +
> +static struct platform_driver qcom_ssc_block_bus_driver = {
> +       .probe = qcom_ssc_block_bus_probe,
> +       .remove = qcom_ssc_block_bus_remove,
> +       .driver = {
> +               .name = "qcom-ssc-block-bus",
> +               .of_match_table = qcom_ssc_block_bus_of_match,
> +       },
> +};
> +
> +module_platform_driver(qcom_ssc_block_bus_driver);
> +
> +MODULE_DESCRIPTION("A driver for handling the init sequence needed for accessing the SSC block on (some) qcom SoCs over AHB");
> +MODULE_AUTHOR("Michael Srba <Michael.Srba@seznam.cz>");
> +MODULE_LICENSE("GPL v2");
> --
> 2.34.1
>

  parent reply	other threads:[~2022-02-27 20:25 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-02-20 21:20 [PATCH v8 1/5] dt-bindings: clock: gcc-msm8998: Add definitions of SSC-related clocks michael.srba
2022-02-20 21:20 ` [PATCH v8 2/5] clk: qcom: gcc-msm8998: add " michael.srba
2022-02-25  0:55   ` Stephen Boyd
2022-02-20 21:20 ` [PATCH v8 3/5] dt-bindings: bus: add device tree bindings for qcom,ssc-block-bus michael.srba
2022-02-27 20:12   ` Jeffrey Hugo
2022-02-20 21:20 ` [PATCH v8 4/5] drivers: bus: add driver for initializing the SSC bus on (some) qcom SoCs michael.srba
2022-02-22 22:29   ` kernel test robot
2022-02-27 20:25   ` Jeffrey Hugo [this message]
2022-02-20 21:20 ` [PATCH v8 5/5] arm64: dts: qcom: msm8998: reserve potentially inaccessible clocks michael.srba
2022-02-27 20:28   ` Jeffrey Hugo

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAOCk7NpXa03fzRSXEEwkBANZNXyb_LLAje1EMsaprUFv0ew2HQ@mail.gmail.com \
    --to=jeffrey.l.hugo@gmail.com \
    --cc=agross@kernel.org \
    --cc=arnd@arndb.de \
    --cc=bjorn.andersson@linaro.org \
    --cc=devicetree@vger.kernel.org \
    --cc=f.fainelli@gmail.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=linus.walleij@linaro.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-clk@vger.kernel.org \
    --cc=michael.srba@seznam.cz \
    --cc=p.zabel@pengutronix.de \
    --cc=robh+dt@kernel.org \
    --cc=saravanak@google.com \
    --cc=sboyd@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.