Hi,
On 4/15/19 2:50 PM, Ard Biesheuvel wrote:
On Mon, 8 Apr 2019 at 17:33, Jeremy Linton <jeremy.lin...@arm.com> wrote:
The mcbin (and likely others) have a nonstandard uart clock.
This means that the earlycon programming will incorrectly set
the baud rate if it is specified. The way around this is to tell
the kernel to continue using the preprogrammed baud rate. This
is done by setting the baud to 0.
Further, the SPCR and DSDT serial port need to match the port
address and port access type for the kernel to conclude they
are the same.
So while ARM_GAS32 is correct for earlycon (it can be used alone
on the kernel command line) by providing the reg-shift=2 value,
it also sets the io type to MMIO32, which doesn't match the DSDT
defined MMIO. This means that the actual console will never appear.
The obvious fix is to set reg-width=4 in DSDT, but that also changes
the accesssors to 32-bits (similarly to earlycon) and results in
console failure.
Given the breakage on the BSDs and Marcin's followup patch, I'd like
to understand why this breaks.
To me, it seems we are conflating register bit width with minimum
access size. However, as far as I can tell, mmio32 sets both the
register width and access size to 32 bits. So is the reason that it
works for earlycon that we only use it for output?
Yes, I went back and looked at the SPCR/earlycon code and that is likely
the core problem. It should be passing both the access width and the
register size. I will look at tossing a patch in the next couple days.
I'm wondering how many of the quirks are in there to work around this
deficiency.
As far as the mcbin, that appears to be the case, just writing the
output shift register at the wrong width (32-bit rather than 8) seems to
be ok (per earlycon), but anything beyond that is a problem.
So the less obvious fix, is to use the GAS8 specifier. This means
that earlycon needs to be fully specified as
earlycon=uart,mmio32,0xf0512000, but has the extremely useful feature
that the console default works without any user interaction.
If in the future marvell decides to define their own ACPI id for the
console and upstream a quirk, the ARM_GAS8 portion of this should
be reverted.
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Jeremy Linton <jeremy.lin...@arm.com>
---
Silicon/Marvell/Armada7k8k/AcpiTables/Spcr.aslc | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/Silicon/Marvell/Armada7k8k/AcpiTables/Spcr.aslc
b/Silicon/Marvell/Armada7k8k/AcpiTables/Spcr.aslc
index e78bb9036f..06c7af069c 100644
--- a/Silicon/Marvell/Armada7k8k/AcpiTables/Spcr.aslc
+++ b/Silicon/Marvell/Armada7k8k/AcpiTables/Spcr.aslc
@@ -30,11 +30,11 @@ EFI_ACPI_SERIAL_PORT_CONSOLE_REDIRECTION_TABLE Spcr = {
{ EFI_ACPI_RESERVED_BYTE,
EFI_ACPI_RESERVED_BYTE,
EFI_ACPI_RESERVED_BYTE }, //
Reserved1[3]
- ARM_GAS32 (FixedPcdGet64(PcdSerialRegisterBase)), //
BaseAddress
+ ARM_GAS8 (FixedPcdGet64(PcdSerialRegisterBase)), //
BaseAddress
EFI_ACPI_SERIAL_PORT_CONSOLE_REDIRECTION_TABLE_INTERRUPT_TYPE_GIC, //
InterruptType
0, // Irq
51, //
GlobalSystemInterrupt
- EFI_ACPI_SERIAL_PORT_CONSOLE_REDIRECTION_TABLE_BAUD_RATE_115200, //
BaudRate
+ 0, //
Keep Firmware Baud
EFI_ACPI_SERIAL_PORT_CONSOLE_REDIRECTION_TABLE_PARITY_NO_PARITY, //
Parity
EFI_ACPI_SERIAL_PORT_CONSOLE_REDIRECTION_TABLE_STOP_BITS_1, //
StopBits
0, //
FlowControl
--
2.20.1
------------
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#38670): https://edk2.groups.io/g/devel/message/38670
Mute This Topic: https://groups.io/mt/30980002/1761188
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [ard.biesheu...@linaro.org]
------------
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#39255): https://edk2.groups.io/g/devel/message/39255
Mute This Topic: https://groups.io/mt/30980002/21656
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-
