Device model by remexre · Pull Request #70 · UMN-Kernel-Object/ukoos

remexre · 2025-11-09T23:24:00Z

On top of #63; merge that one first, then rebase this.

Adds initializers. These are used to register drivers with the appropriate subsystems.
Adds the device model and documentation about it.
Adds support for enumerating devices in the Devicetree.
Adds the UART device class.
Adds an incomplete ns16550a driver. This is the UART we get in QEMU, and most UARTs in the wild are compatible with it. This driver is currently very basic, since we don't have interrupts (Trap handling #9) and it doesn't do DMA.

Relevant to #10, but it is not resolved by this PR.

The UART device class and ns16550a driver here are not very finished.
They currently act as an example of the device model, but don't implement all of what's needed for the driver itself.

Signed-off-by: Amy Ringo <me@remexre.com>

These functions are for mapping IO memory. Signed-off-by: Amy Ringo <me@remexre.com>

Signed-off-by: Amy Ringo <me@remexre.com>

remexre · 2026-01-13T00:28:47Z

Since there are other drivers in development (#81), despite the actual ns16550a driver and UART device class here being very unfinished, I'm un-drafting this and looking for review.

KieranMusser · 2026-01-13T05:22:11Z

I'll look a bit more into this later (as I rewrite my driver), but should we include PCI somewhere in this? I feel it might be good to separate the PCI driver from the RTL8139 driver. I'll be doing all the work for PCI as a part of the RTL fixing, but PCI is useful to more than just that one driver.

remexre · 2026-01-13T05:30:32Z

I'll look a bit more into this later (as I rewrite my driver), but should we include PCI somewhere in this? I feel it might be good to separate the PCI driver from the RTL8139 driver.

Depends how big it is, I suppose; if the PCI and RTL8139 drivers are small enough together that they make sense to review together, then eh, whatever's fine.

If the PCI driver is large enough to need separate review, I would make a separate PR for it that has a target branch of this one, then set the RTL8139 PR to have a target branch of that one.

A few advantages there:

The GitHub UI only shows the diff against that other branch, so your reviewers don't end up re-reviewing commits that're really from the other branch.
The list of PRs makes it visually obvious when a branch depends on another in this way.
Once the target branch gets merged, GitHub will email you and your reviewer(s). The message is phrased a bit rudely ("dismissed their stale review"), but... it's at least a ping lol.

jastintime · 2026-02-15T05:48:48Z

src/kernel/include/device.h

+  struct list_head parent_children;
+
+  /**
+   * The children of the node.


this is an unnecessary comment.

jastintime · 2026-02-15T06:00:29Z

src/kernel/include/device.h

+ * Registers a device in the tree of devices and with its device class.
+ *
+ * - `name` must be filled out.
+ * - `parent` and `device_class` must be `nullptr`.


im confused, the attribute says nonnull but the comment says must be nullptr.

jastintime · 2026-02-15T06:03:22Z

doc/kernel/devices.md

+
+## Device classes
+
+Each device probably provides some functionality other than simply existing.


this is probably not necessary.

jastintime · 2026-02-15T06:04:08Z

doc/kernel/devices.md

+
+Typically, a device class consists of three things: a vtable struct, a device struct, and a list of devices.
+
+The vtable is simply a struct with methods for that kind of device.


jastintime · 2026-02-15T06:19:22Z

src/kernel/arch/riscv64/paging.c

+
+  uaddr lo, hi;
+  struct vma *vma;
+  vma = vma_alloc(&kernel_virtual_allocator, size >> 12);


size >> 12 is some magic

jastintime · 2026-02-15T06:27:08Z

src/kernel/drivers/ns16550a.c

+
+/**
+ * The bits of the Interrupt Enable Register.
+ */


can we add someone a link to the spec sheet for this UART device.

im unable to verify these cant find the spec sheet.

jastintime · 2026-02-15T06:40:29Z

src/kernel/drivers/ns16550a.c

+  return &device->device;
+
+fail:
+  if (device) {


might be a good idea to have a device_free(), so we don't have to remember to free the format name, unmap and free the whole thing.

jastintime · 2026-02-15T06:43:25Z

src/kernel/include/devices/uart.h

+  const struct uart_ops *ops;
+
+  /**
+   * A pointer to the _same_ device.


what does same device mean here.

jastintime · 2026-02-15T06:44:53Z

src/kernel/include/mm/paging.h


+/**
+ * Maps a region of memory for memory-mapped IO. This may involve special flags
+ * or other setup to prevent accesses from being cached.


where would this setup be? or do we not have that option yet?

jastintime · 2026-02-15T07:05:20Z

src/kernel/devicetree.c

+
+  if (!handlers) {
+    assert(!handlers_len);
+    handlers_cap = 16;


magic value

remexre mentioned this pull request Nov 9, 2025

snps,dw-apb-uart driver #71

Open

remexre changed the base branch from trunk to remexre/split-up-objects-in-makefile January 3, 2026 01:18

remexre force-pushed the remexre/split-up-objects-in-makefile branch 2 times, most recently from 1d3bead to ac9d91b Compare January 12, 2026 15:35

remexre force-pushed the remexre/driver-model branch from 62eddfa to 45e23c0 Compare January 12, 2026 15:39

remexre force-pushed the remexre/split-up-objects-in-makefile branch from ac9d91b to 18dd63d Compare January 12, 2026 16:04

remexre force-pushed the remexre/driver-model branch from 45e23c0 to 628e825 Compare January 12, 2026 16:04

remexre force-pushed the remexre/split-up-objects-in-makefile branch from 18dd63d to b3dda97 Compare January 12, 2026 23:20

Base automatically changed from remexre/split-up-objects-in-makefile to trunk January 12, 2026 23:23

remexre mentioned this pull request Jan 13, 2026

Implement RTL8139 Driver #81

Draft

remexre added 8 commits January 12, 2026 18:17

Adds support for pointers to align_down and align_up.

7b070bf

Signed-off-by: Amy Ringo <me@remexre.com>

Implements iomem_map and iomem_unmap.

05b6d9b

These functions are for mapping IO memory. Signed-off-by: Amy Ringo <me@remexre.com>

Makes devicetree_print prints the path to the node.

d73b0e0

Signed-off-by: Amy Ringo <me@remexre.com>

Adds initializers and documentation about them.

c352b56

Signed-off-by: Amy Ringo <me@remexre.com>

Adds a notion of device objects.

69ffac2

Signed-off-by: Amy Ringo <me@remexre.com>

Adds support for enumerating devices in a Devicetree.

3c20f27

Signed-off-by: Amy Ringo <me@remexre.com>

Adds the UART device class.

fae6a32

Signed-off-by: Amy Ringo <me@remexre.com>

Adds the ns16550a driver.

ff452dc

Signed-off-by: Amy Ringo <me@remexre.com>

remexre force-pushed the remexre/driver-model branch from 628e825 to ff452dc Compare January 13, 2026 00:17

remexre marked this pull request as ready for review January 13, 2026 00:29

remexre mentioned this pull request Feb 14, 2026

Adds GCC 15 support #91

Merged

jastintime requested changes Feb 15, 2026

View reviewed changes


		## Device classes

		Each device probably provides some functionality other than simply existing.


		Typically, a device class consists of three things: a vtable struct, a device struct, and a list of devices.

		The vtable is simply a struct with methods for that kind of device.

Conversation

remexre commented Nov 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

remexre commented Jan 13, 2026

Uh oh!

KieranMusser commented Jan 13, 2026

Uh oh!

remexre commented Jan 13, 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!

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!

3 participants

Comments

remexre commented Nov 9, 2025 •

edited

Loading