Hi,
On 04/09/2024 20:07, Mukesh Kumar Savaliya wrote:
Thanks Neil !
On 8/30/2024 1:17 PM, neil.armstrong@xxxxxxxxxx wrote:
Hi,
On 29/08/2024 11:24, Mukesh Kumar Savaliya wrote:
This Series adds support to share QUP based I2C SE between subsystems.
Each subsystem should have its own GPII which interacts between SE and
GSI DMA HW engine.
Subsystem must acquire Lock over the SE on GPII channel so that it
gets uninterrupted control till it unlocks the SE. It also makes sure
the commonly shared TLMM GPIOs are not touched which can impact other
subsystem or cause any interruption. Generally, GPIOs are being
unconfigured during suspend time.
GSI DMA engine is capable to perform requested transfer operations
from any of the SE in a seamless way and its transparent to the
subsystems. Make sure to enable “qcom,shared-se” flag only while
enabling this feature. I2C client should add in its respective parent
node.
---
Mukesh Kumar Savaliya (4):
dt-bindindgs: i2c: qcom,i2c-geni: Document shared flag
dma: gpi: Add Lock and Unlock TRE support to access SE exclusively
soc: qcom: geni-se: Export function geni_se_clks_off()
i2c: i2c-qcom-geni: Enable i2c controller sharing between two
subsystems
.../bindings/i2c/qcom,i2c-geni-qcom.yaml | 4 ++
drivers/dma/qcom/gpi.c | 37 ++++++++++++++++++-
drivers/i2c/busses/i2c-qcom-geni.c | 29 +++++++++++----
drivers/soc/qcom/qcom-geni-se.c | 4 +-
include/linux/dma/qcom-gpi-dma.h | 6 +++
include/linux/soc/qcom/geni-se.h | 3 ++
6 files changed, 74 insertions(+), 9 deletions(-)
I see in downstream that this flag is used on the SM8650 qupv3_se6_i2c,
and that on the SM8650-HDK this i2c is shared between the aDSP battmgr and
the linux to access the HDMI controller.
Is this is the target use-case ?
Not exactly that usecase. Here making it generic in a way to transfer data which is pushed from two subsystems independently. Consider for example one is ADSP i2c client and another is Linux i2c client. Not sure in what manner battmgr and HDMI sends traffic. we can debug it separately over that email.
Considering battmgr runs in ADSP, it matches this use-case, no ?
We have some issues on this platform that crashes the system when Linux
does some I2C transfers while battmgr does some access at the same time,
the problem is that on the Linux side the i2c uses the SE DMA and not GPI
because fifo_disable=0 so by default this bypasses GPI.
A temporary fix has been merged:
https://lore.kernel.org/all/20240605-topic-sm8650-upstream-hdk-iommu-fix-v1-1-9fd7233725fa@xxxxxxxxxx/
but it's clearly not a real solution
Seems you have added SID for the GPII being used from linux side. Need to know why you have added it and is it helping ? I have sent an email to know more about this issue before 2 weeks.
I've added this because it actually avoids crashing when doing I2C6 transactions over SE DMA, now we need to understand why.
What would be the solution to use the shared i2c with on one side battmgr
using GPI and the kernel using SE DMA ?
I have already sent an email on this issue, please respond on it. We shall debug it separately since this feature about sharing is still under implementation as you know about this patch series.
Sorry for the delay, I was technically unable to answer, let me resume it now that I'm able again.
Thanks,
Neil
In this case, shouldn't we force using GPI on linux with:
==============><=====================================================================
diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
index ee2e431601a6..a15825ea56de 100644
--- a/drivers/i2c/busses/i2c-qcom-geni.c
+++ b/drivers/i2c/busses/i2c-qcom-geni.c
@@ -885,7 +885,7 @@ static int geni_i2c_probe(struct platform_device *pdev)
else
fifo_disable = readl_relaxed(gi2c->se.base + GENI_IF_DISABLE_RO) & FIFO_IF_DISABLE;
- if (fifo_disable) {
+ if (gi2c->is_shared || fifo_disable) {
/* FIFO is disabled, so we can only use GPI DMA */
gi2c->gpi_mode = true;
ret = setup_gpi_dma(gi2c);
==============><=====================================================================
Neil
