| Message ID | cover.1539705285.git.leonard.crestez@nxp.com |
|---|---|
| Headers | show |
| Series | Port mxs-dcp to imx6ull and imx6sll | expand |
Hi Leonard, On Tue, Oct 16, 2018 at 12:58 PM Leonard Crestez <leonard.crestez@nxp.com> wrote: > + /* DCP clock is optional, only used on some SOCs */ > + sdcp->dcp_clk = devm_clk_get(dev, "dcp"); > + if (IS_ERR(sdcp->dcp_clk)) { > + if (sdcp->dcp_clk != ERR_PTR(-ENOENT)) > + return PTR_ERR(sdcp->dcp_clk); > + sdcp->dcp_clk = NULL; This dcp_clk assignment to NULL does not seem to be necessary. > + > + ret = clk_prepare_enable(sdcp->dcp_clk); > + if (ret) > + return ret; > > ret = devm_request_irq(dev, dcp_vmi_irq, mxs_dcp_irq, 0, > "dcp-vmi-irq", sdcp); > if (ret) { > dev_err(dev, "Failed to claim DCP VMI IRQ!\n"); In case of subsequent errors you should call clk_disable_unprepare(sdcp->dcp_clk).
On Tue, Oct 16, 2018 at 12:59 PM Leonard Crestez <leonard.crestez@nxp.com> wrote: > > The DCP block on 6ull has no major differences other than requiring > explicit clock enabling. > > Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com> Reviewed-by: Fabio Estevam <festevam@gmail.com>
On Tue, 2018-10-16 at 13:03 -0300, Fabio Estevam wrote: > > + /* DCP clock is optional, only used on some SOCs */ > > + sdcp->dcp_clk = devm_clk_get(dev, "dcp"); > > + if (IS_ERR(sdcp->dcp_clk)) { > > + if (sdcp->dcp_clk != ERR_PTR(-ENOENT)) > > + return PTR_ERR(sdcp->dcp_clk); > > + sdcp->dcp_clk = NULL; > > This dcp_clk assignment to NULL does not seem to be necessary. The clk API ignores all calls if clk == NULL (which is very nice) but not if IS_ERR(clk). Setting dcp_clk to NULL avoids sprinkling other checks. > > + ret = clk_prepare_enable(sdcp->dcp_clk); > > + if (ret) > > + return ret; > > > > ret = devm_request_irq(dev, dcp_vmi_irq, mxs_dcp_irq, 0, > > "dcp-vmi-irq", sdcp); > > if (ret) { > > dev_err(dev, "Failed to claim DCP VMI IRQ!\n"); > > In case of subsequent errors you should call > clk_disable_unprepare(sdcp->dcp_clk). Yes, will resend. Maybe some day clk consumer will automatically undo pending prepare/enables on release?
The DCP block is present on 6sll and 6ull but not enabled. The hardware is mostly compatible with 6sl, the only important difference is that explicit clock enabling is required. There were several issues with the functionality of this driver (it didn't even probe properly) but they are fixed in cryptodev/master by this series: https://lore.kernel.org/patchwork/cover/994874/ --- Changes since v2: * Rename 6ull dcp node to crypto@* * Use clk_prepare_enable on probe * Fix not unpreparing on remove * Realign dcp_clk struct member with rest of attributes (using tabs) * Link to v2: https://lore.kernel.org/patchwork/cover/999795/ Changes since v1: * Add devicetree maintainers for dt-bindings * Add a patch enabling in imx_v6_v7_defconfig. Since tcrypt now passes this shouldn't cause any issues * Link to v1: https://lore.kernel.org/patchwork/cover/994893/ Leonard Crestez (4): dt-bindings: crypto: Mention clocks for mxs-dcp crypto: mxs-dcp - Add support for dcp clk ARM: dts: imx6ull: Add dcp node ARM: imx_v6_v7_defconfig: Enable CRYPTO_DEV_MXS_DCP .../devicetree/bindings/crypto/fsl-dcp.txt | 2 ++ arch/arm/boot/dts/imx6ull.dtsi | 10 ++++++++++ arch/arm/configs/imx_v6_v7_defconfig | 1 + drivers/crypto/mxs-dcp.c | 15 +++++++++++++++ 4 files changed, 28 insertions(+)