* [PATCH 0/2] riscv/clk: sophgo: Fix EFI RTC on Milk-V Pioneer
@ 2026-11-21 10:38 Han Gao via B4 Relay
2026-11-21 10:38 ` [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes Han Gao via B4 Relay
2026-11-21 10:38 ` [PATCH 2/2] clk: sophgo: sg2042-clkgen: Keep gates of reserved devices enabled Han Gao via B4 Relay
0 siblings, 2 replies; 5+ messages in thread
From: Han Gao via B4 Relay @ 2026-11-21 10:38 UTC (permalink / raw)
To: Ard Biesheuvel, Ilias Apalodimas, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Alexandre Ghiti, Atish Patra, Stephen Boyd,
Brian Masney, Jerome Brunet, Chen Wang, Inochi Amaoto
Cc: Han Gao, linux-efi, linux-riscv, linux-kernel, linux-clk, sophgo,
Han Gao
Pioneer firmware can take over the DS1307 on I2C0 and expose it through
EFI runtime services, so Linux uses rtc-efi. I2C0 is then marked
"reserved" in the device tree.
Two things break GetTime() in that setup:
- RISC-V maps EFI runtime MMIO as normal memory, and firmware takes a
load access fault on the clock controller. Patch 1 maps it as I/O.
- I2C0 shares a clock gate with the other I2C controllers. Once Linux
runtime-suspends them, the gate goes off and GetTime() returns
EFI_NOT_READY. Patch 2 keeps gates used by reserved devices on.
Tested on Milk-V Pioneer.
Signed-off-by: Han Gao <gaohan@iscas.ac.cn>
---
Han Gao (2):
riscv: efi: Map runtime MMIO with device attributes
clk: sophgo: sg2042-clkgen: Keep gates of reserved devices enabled
arch/riscv/kernel/efi.c | 2 +-
drivers/clk/sophgo/clk-sg2042-clkgen.c | 38 +++++++++++++++++++++++++++++++++-
2 files changed, 38 insertions(+), 2 deletions(-)
---
base-commit: aebb7274476403040d582c61478df8f4a7bf9cd3
change-id: 20261121-b4-sg2042-efi-rtc-e4d5222513ae
Best regards,
--
Han Gao <gaohan@iscas.ac.cn>
^ permalink raw reply [flat|nested] 5+ messages in thread* [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes 2026-11-21 10:38 [PATCH 0/2] riscv/clk: sophgo: Fix EFI RTC on Milk-V Pioneer Han Gao via B4 Relay @ 2026-11-21 10:38 ` Han Gao via B4 Relay 2026-11-21 12:42 ` Ard Biesheuvel 2026-11-21 10:38 ` [PATCH 2/2] clk: sophgo: sg2042-clkgen: Keep gates of reserved devices enabled Han Gao via B4 Relay 1 sibling, 1 reply; 5+ messages in thread From: Han Gao via B4 Relay @ 2026-11-21 10:38 UTC (permalink / raw) To: Ard Biesheuvel, Ilias Apalodimas, Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti, Atish Patra, Stephen Boyd, Brian Masney, Jerome Brunet, Chen Wang, Inochi Amaoto Cc: Han Gao, linux-efi, linux-riscv, linux-kernel, linux-clk, sophgo, Han Gao From: Han Gao <gaohan@iscas.ac.cn> EFI_MEMORY_MAPPED_IO regions are mapped with PAGE_KERNEL, so device registers end up mapped as normal memory. On Milk-V Pioneer, GetTime() reads the clock controller registers and faults: Oops - load access fault [#1] Workqueue: efi_runtime efi_call_rts status: 0000000200000120 badaddr: 0000007030012000 cause: 0000000000000005 Use PAGE_KERNEL_IO for these regions. Fixes: b91540d52a08 ("RISC-V: Add EFI runtime services") Signed-off-by: Han Gao <gaohan@iscas.ac.cn> --- arch/riscv/kernel/efi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/riscv/kernel/efi.c b/arch/riscv/kernel/efi.c index 2d3cc57b4535..11e85d0ebc91 100644 --- a/arch/riscv/kernel/efi.c +++ b/arch/riscv/kernel/efi.c @@ -22,7 +22,7 @@ static __init pgprot_t efimem_to_pgprot_map(efi_memory_desc_t *md) u32 type = md->type; if (type == EFI_MEMORY_MAPPED_IO) - return PAGE_KERNEL; + return PAGE_KERNEL_IO; /* R-- */ if ((attr & (EFI_MEMORY_XP | EFI_MEMORY_RO)) == -- 2.55.0 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes 2026-11-21 10:38 ` [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes Han Gao via B4 Relay @ 2026-11-21 12:42 ` Ard Biesheuvel 2026-11-21 14:48 ` Han Gao 0 siblings, 1 reply; 5+ messages in thread From: Ard Biesheuvel @ 2026-11-21 12:42 UTC (permalink / raw) To: gaohan, Ilias Apalodimas, Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti, Atish Patra, Stephen Boyd, Brian Masney, Jerome Brunet, Chen Wang, Inochi Amaoto Cc: Han Gao, linux-efi, linux-riscv, linux-kernel, linux-clk, sophgo On Sat, 21 Nov 2026, at 11:38, Han Gao via B4 Relay wrote: > From: Han Gao <gaohan@iscas.ac.cn> > > EFI_MEMORY_MAPPED_IO regions are mapped with PAGE_KERNEL, so device > registers end up mapped as normal memory. > > On Milk-V Pioneer, GetTime() reads the clock controller registers and > faults: > > Oops - load access fault [#1] > Workqueue: efi_runtime efi_call_rts > status: 0000000200000120 badaddr: 0000007030012000 cause: 0000000000000005 > > Use PAGE_KERNEL_IO for these regions. > > Fixes: b91540d52a08 ("RISC-V: Add EFI runtime services") > Signed-off-by: Han Gao <gaohan@iscas.ac.cn> > --- > arch/riscv/kernel/efi.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > Acked-by: Ard Biesheuvel <ardb@kernel.org> I can take this as a fix via the EFI tree if you like. > diff --git a/arch/riscv/kernel/efi.c b/arch/riscv/kernel/efi.c > index 2d3cc57b4535..11e85d0ebc91 100644 > --- a/arch/riscv/kernel/efi.c > +++ b/arch/riscv/kernel/efi.c > @@ -22,7 +22,7 @@ static __init pgprot_t > efimem_to_pgprot_map(efi_memory_desc_t *md) > u32 type = md->type; > > if (type == EFI_MEMORY_MAPPED_IO) > - return PAGE_KERNEL; > + return PAGE_KERNEL_IO; > > /* R-- */ > if ((attr & (EFI_MEMORY_XP | EFI_MEMORY_RO)) == > > -- > 2.55.0 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes 2026-11-21 12:42 ` Ard Biesheuvel @ 2026-11-21 14:48 ` Han Gao 0 siblings, 0 replies; 5+ messages in thread From: Han Gao @ 2026-11-21 14:48 UTC (permalink / raw) To: Ard Biesheuvel Cc: Han Gao, gaohan, Ilias Apalodimas, Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti, Atish Patra, Stephen Boyd, Brian Masney, Jerome Brunet, Chen Wang, Inochi Amaoto, linux-efi, linux-riscv, linux-kernel, linux-clk, sophgo On Sat, 21 Nov 2026 13:42:20 +0100 "Ard Biesheuvel" <ardb@kernel.org> wrote: > > > On Sat, 21 Nov 2026, at 11:38, Han Gao via B4 Relay wrote: > > From: Han Gao <gaohan@iscas.ac.cn> > > > > EFI_MEMORY_MAPPED_IO regions are mapped with PAGE_KERNEL, so device > > registers end up mapped as normal memory. > > > > On Milk-V Pioneer, GetTime() reads the clock controller registers and > > faults: > > > > Oops - load access fault [#1] > > Workqueue: efi_runtime efi_call_rts > > status: 0000000200000120 badaddr: 0000007030012000 cause: 0000000000000005 > > > > Use PAGE_KERNEL_IO for these regions. > > > > Fixes: b91540d52a08 ("RISC-V: Add EFI runtime services") > > Signed-off-by: Han Gao <gaohan@iscas.ac.cn> > > --- > > arch/riscv/kernel/efi.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > Acked-by: Ard Biesheuvel <ardb@kernel.org> > > I can take this as a fix via the EFI tree if you like. Yes, please take it through the EFI tree. Thanks! > > > > diff --git a/arch/riscv/kernel/efi.c b/arch/riscv/kernel/efi.c > > index 2d3cc57b4535..11e85d0ebc91 100644 > > --- a/arch/riscv/kernel/efi.c > > +++ b/arch/riscv/kernel/efi.c > > @@ -22,7 +22,7 @@ static __init pgprot_t > > efimem_to_pgprot_map(efi_memory_desc_t *md) > > u32 type = md->type; > > > > if (type == EFI_MEMORY_MAPPED_IO) > > - return PAGE_KERNEL; > > + return PAGE_KERNEL_IO; > > > > /* R-- */ > > if ((attr & (EFI_MEMORY_XP | EFI_MEMORY_RO)) == > > > > -- > > 2.55.0 > > _______________________________________________ > linux-riscv mailing list > linux-riscv@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-riscv Sent using hkml (https://github.com/sjp38/hackermail) ^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 2/2] clk: sophgo: sg2042-clkgen: Keep gates of reserved devices enabled 2026-11-21 10:38 [PATCH 0/2] riscv/clk: sophgo: Fix EFI RTC on Milk-V Pioneer Han Gao via B4 Relay 2026-11-21 10:38 ` [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes Han Gao via B4 Relay @ 2026-11-21 10:38 ` Han Gao via B4 Relay 1 sibling, 0 replies; 5+ messages in thread From: Han Gao via B4 Relay @ 2026-11-21 10:38 UTC (permalink / raw) To: Ard Biesheuvel, Ilias Apalodimas, Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti, Atish Patra, Stephen Boyd, Brian Masney, Jerome Brunet, Chen Wang, Inochi Amaoto Cc: Han Gao, linux-efi, linux-riscv, linux-kernel, linux-clk, sophgo, Han Gao From: Han Gao <gaohan@iscas.ac.cn> On Milk-V Pioneer, firmware can own I2C0 to back the EFI RTC, with the node marked status = "reserved". I2C0 shares GATE_CLK_APB_I2C with the I2C controllers Linux uses. When those runtime-suspend, the gate is switched off under the firmware and GetTime() returns EFI_NOT_READY. Walk the reserved children of /soc and match their clock references, as done in commit 6aa175476490 ("clk: renesas: cpg-mssr: Ignore all clocks assigned to non-Linux system"). Unlike that driver, use CLK_IS_CRITICAL rather than CLK_IGNORE_UNUSED: the gate is shared with devices Linux drives, and their runtime PM would still turn it off. Signed-off-by: Han Gao <gaohan@iscas.ac.cn> --- drivers/clk/sophgo/clk-sg2042-clkgen.c | 38 +++++++++++++++++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-) diff --git a/drivers/clk/sophgo/clk-sg2042-clkgen.c b/drivers/clk/sophgo/clk-sg2042-clkgen.c index 6abc2c973c49..414904965edf 100644 --- a/drivers/clk/sophgo/clk-sg2042-clkgen.c +++ b/drivers/clk/sophgo/clk-sg2042-clkgen.c @@ -11,6 +11,8 @@ #include <linux/clk.h> #include <linux/clk-provider.h> #include <linux/io.h> +#include <linux/of.h> +#include <linux/of_clk.h> #include <linux/platform_device.h> #include <asm/div64.h> @@ -840,6 +842,35 @@ static int sg2042_clk_register_divs(struct device *dev, return ret; } +static bool sg2042_clk_is_reserved(struct device_node *clkgen, unsigned int id) +{ + struct device_node *soc __free(device_node) = of_find_node_by_path("/soc"); + struct device_node *node; + + for_each_reserved_child_of_node(soc, node) { + unsigned int i, count = of_clk_get_parent_count(node); + + for (i = 0; i < count; i++) { + struct of_phandle_args clkspec; + bool match; + + if (of_parse_phandle_with_args(node, "clocks", "#clock-cells", + i, &clkspec)) + continue; + + match = clkspec.np == clkgen && clkspec.args_count == 1 && + clkspec.args[0] == id; + of_node_put(clkspec.np); + if (match) { + of_node_put(node); + return true; + } + } + } + + return false; +} + static int sg2042_clk_register_gates(struct device *dev, struct sg2042_clk_data *clk_data, const struct sg2042_gate_clock gate_clks[], @@ -850,7 +881,12 @@ static int sg2042_clk_register_gates(struct device *dev, int i, ret = 0; for (i = 0; i < num_gate_clks; i++) { + unsigned long flags; + gate = &gate_clks[i]; + flags = gate->hw.init->flags; + if (sg2042_clk_is_reserved(dev->of_node, gate->id)) + flags |= CLK_IS_CRITICAL; hw = __devm_clk_hw_register_gate (dev, NULL, @@ -858,7 +894,7 @@ static int sg2042_clk_register_gates(struct device *dev, NULL, gate->hw.init->parent_hws[0], NULL, - gate->hw.init->flags, + flags, clk_data->iobase + gate->offset_enable, gate->bit_idx, 0, -- 2.55.0 ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2026-11-21 14:48 UTC | newest] Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2026-11-21 10:38 [PATCH 0/2] riscv/clk: sophgo: Fix EFI RTC on Milk-V Pioneer Han Gao via B4 Relay 2026-11-21 10:38 ` [PATCH 1/2] riscv: efi: Map runtime MMIO with device attributes Han Gao via B4 Relay 2026-11-21 12:42 ` Ard Biesheuvel 2026-11-21 14:48 ` Han Gao 2026-11-21 10:38 ` [PATCH 2/2] clk: sophgo: sg2042-clkgen: Keep gates of reserved devices enabled Han Gao via B4 Relay
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®