PMP: initialization and experiments by WhatAmISupposedToPutHere · Pull Request #545 · AsahiLinux/m1n1

WhatAmISupposedToPutHere · 2026-02-25T23:03:22Z

The thingy that makes your mac use less power.

jannau · 2026-02-26T07:52:19Z

src/kboot.c

+
+    ADT_FOREACH_PROPERTY(adt, pmp_iop_anode, prop)
+    {
+        if (fdt_setprop(dt, tunables_node, prop->name, prop->value, prop->size))


I suspect we want to filter this at least lightly. compatible = [iop-nub,rtbuddy-v2] and AAPL,phandle shouldn't be copied. There are further properties an OSS driver probably does not need.
I hope "these are the property names apple uses in the ADT" is a convincing argument during dt-bindings / driver review but don't really see a sane alternative.

The annoying part is that firmware requests the property values from this node by name, so i don't even have a static list of properties that it needs. Can only treat them as an opaque key/value dict.

does it request "soc-device" / "controller" / "lts-config"? looking at the larger properties.
I'd expected "region-", "segment-", "pre-loaded", "firmware-name", "dram-capacity", "coredump-enable" can be removed as well. Primarily looking at properties which could end up confusing

the big three are requested, but the others on the list can be excluded

jannau · 2026-02-26T08:00:24Z

proxyclient/experiments/pmp_init.py

+        pio_base = u.adt["/arm-io/dart-pmp"].pio_vm_base
+        granularity = u.adt["/arm-io/dart-pmp"].pio_granularity
+        i = 0
+        for (host_addr, size) in [(0x282000000, 0x1000000),(0x304000000, 0x1000000),(0x383000000, 0x1000000),(0x402000000, 0x1000000),(0x210e70000, 0x90000),(0x211e70000, 0x90000),(0x212e70000, 0x90000)]:


Are these address/size pairs observed by tracing? On which device? Can we derive them from the ADT?

that is on a m1 pro, they are derived from adt (reg 4 and following, i have written them down in linux driver's dt)

can we documents that this is only expected to run on t6000 (and maybe other t600x or t602x devices)? An compatibility check with early exit would be enough. Please mention it in the commit message as well.

Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

src/kboot.c

jannau · 2026-02-28T10:39:58Z

src/kboot.c

+
+    ADT_FOREACH_PROPERTY(adt, pmp_iop_anode, prop)
+    {
+        if (fdt_setprop(dt, tunables_node, prop->name, prop->value, prop->size))


does it request "soc-device" / "controller" / "lts-config"? looking at the larger properties.
I'd expected "region-", "segment-", "pre-loaded", "firmware-name", "dram-capacity", "coredump-enable" can be removed as well. Primarily looking at properties which could end up confusing

jannau

The commits are missing Signed-off-by:

src/kboot.c

jannau · 2026-03-12T20:59:24Z

src/kboot.c

+            bail("FDT: failed to transfer pmp tunable");
+    next_adt_prop:;
+    }
+    return 0;


please update pmp_node's "status" to "okay". I will disable the pmp node in the linux device trees so we can merge the kernel bits before the next m1n1 release

I just realized that this isn't possible due to the power-domains dependencies. This has to be handled in the driver. I don't remember if this is already handled.

I think it should be fine. pmp itself will stay disabled, while the kernel harmlessly writes to pmp_reporting (a shared sram area). Just like macos does, we do not wait for pmp acknowledgements until it has booted

yes, I think so too after looking at it. So I will disable pmp in the kernel DT.

This reverts commit f41676c. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

This reverts commit bfe4027. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

This reverts commit 4be960c. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

src/kboot.c

Copy all the things that PMP needs in order to boot. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

jannau · 2026-03-20T09:08:02Z

src/kboot.c

+    if (ADT_GETPROP(adt, pmp_iop_anode, "dram-capacity", &dram_capacity) < 0)
+        bail("ADT: failed to get dram capacity\n");
+
+    int pmp_node = fdt_path_offset(dt, "pmp");


please move this to the beginning of the function. The j504 (M3 macbook pro) does not have dram-capacity in /arm-io/pmp/iop-pmp-nub.
Checking the FDT first for pmp avoids to fail on new devices with different ADT properties.

jannau · 2026-03-20T09:09:30Z

src/kboot.c

+        bail("ADT: failed to get board id\n");
+    if (ADT_GETPROP(adt, chosen_anode, "dram-vendor-id", &dram_vendor_id) < 0)
+        bail("ADT: failed to get dram vendor id\n");
+    if (ADT_GETPROP(adt, pmp_iop_anode, "dram-capacity", &dram_capacity) < 0)


can we use /chosen/dram-size instead?

/chosen/dram-size from the ADT? Don't think so, whatever this field specifies, i haven't seen an obvious relationship with the system ram size, as both my 32gb and 8gb machines have it set to 8

All device ADTs I could quickly grep have dram-capacity = 8. I just assumed it is the RAM size and only verified that dram-size matches the installed RAM. Leave it as it is then and we'll have to adjust once we support PMP on M3

WhatAmISupposedToPutHere requested review from jannau and svenpeter42 as code owners February 25, 2026 23:03

jannau reviewed Feb 26, 2026

View reviewed changes

WhatAmISupposedToPutHere force-pushed the feldspar branch 2 times, most recently from 973ac05 to c2b0c93 Compare February 26, 2026 11:40

proxyclient: Add PMP experiments

92a9b14

Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

WhatAmISupposedToPutHere force-pushed the feldspar branch from c2b0c93 to 128ade7 Compare February 28, 2026 10:09

jannau reviewed Feb 28, 2026

View reviewed changes

WhatAmISupposedToPutHere force-pushed the feldspar branch from 128ade7 to 5aecef7 Compare February 28, 2026 12:08

jannau reviewed Mar 12, 2026

View reviewed changes

WhatAmISupposedToPutHere added 3 commits March 12, 2026 22:08

Revert "adt: drop unused C macros"

5bfccc5

This reverts commit f41676c. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

Revert "adt: drop adt_next_property_offset()"

492a991

This reverts commit bfe4027. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

Revert "adt: drop ADT_FOREACH_PROPERTY"

aad70d5

This reverts commit 4be960c. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

WhatAmISupposedToPutHere force-pushed the feldspar branch from 5aecef7 to 5cf3eb3 Compare March 12, 2026 21:23

jannau reviewed Mar 13, 2026

View reviewed changes

src/kboot.c Outdated Show resolved Hide resolved

kboot: pmp initialization

2b75771

Copy all the things that PMP needs in order to boot. Signed-off-by: Sasha Finkelstein <fnkl.kernel@gmail.com>

WhatAmISupposedToPutHere force-pushed the feldspar branch from 5cf3eb3 to 2b75771 Compare March 13, 2026 16:14

jannau reviewed Mar 20, 2026

View reviewed changes

Uh oh!

Conversation

WhatAmISupposedToPutHere commented Feb 25, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jannau left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants