Upstream: U-Boot DRAM sizing for sun9i #28

New issue

Open

opened 2026-08-27 23:36:38 +00:00 by tiagoagueda · 1 comment

tiagoagueda commented

2026-08-27 23:36:38 +00:00

Owner

U-Boot reports 2 GiB on a board with 4 GiB fitted; three separate bugs were involved and the
board now correctly sees 3.5 GiB. See 22-dram-4gb.md.

Independent of every kernel series here — different project, different list — so it can go
whenever it is ready.

Tree: u-boot/u-boot
List: u-boot@lists.denx.de (sunxi custodian)
Tooling: patman rather than b4

⚠️ Nothing has been submitted. See 42-upstreaming.md for routing,
order, mechanics, and the DCO / AI-assistance caveats. Patches are exported to patches/,
which is not the same as proposing them.

U-Boot reports 2 GiB on a board with 4 GiB fitted; three separate bugs were involved and the board now correctly sees 3.5 GiB. See [22-dram-4gb.md](22-dram-4gb.md). Independent of every kernel series here — different project, different list — so it can go whenever it is ready. - Tree: u-boot/u-boot - List: u-boot@lists.denx.de (sunxi custodian) - Tooling: `patman` rather than `b4` --- ⚠️ **Nothing has been submitted.** See [42-upstreaming.md](42-upstreaming.md) for routing, order, mechanics, and the DCO / AI-assistance caveats. Patches are exported to `patches/`, which is not the same as proposing them.

tiagoagueda added the

upstream

P3-normal

labels

2026-08-27 23:36:38 +00:00

tiagoagueda referenced this issue

2026-08-27 23:36:39 +00:00

Upstream the accumulated sun9i fixes #20

tiagoagueda referenced this issue

2026-08-28 16:38:22 +00:00

A15 cluster: board hard-resets under any sustained A15 load, at ~80 C #53

tiagoagueda commented

2026-08-29 07:41:10 +00:00

Author

Owner

Unblocked 2026-08-29. #38 was the blocker on this and is now done: a full-range mtest
over 3503 MiB passed with 0 errors, and bdinfo states the geometry independently as one
0xe0000000 bank.

So the series can now say what was actually verified rather than implying it. Suggested
wording for the commit message, since "Linux boots and reports the right size" was the weak
part:

Verified on a Tronsmart Draco AW80 Telos with a full-range U-Boot mtest
(CONFIG_SYS_ALT_MEMTEST, address-line walk included) across 3503 MiB - everything below
U-Boot's own reservation - with zero errors.

One caveat worth carrying: the top 81 MiB cannot be tested from U-Boot because U-Boot lives
there. It is a small fraction and the address-line walk covers the full span either way, but
it should be stated rather than glossed.

**Unblocked 2026-08-29.** #38 was the blocker on this and is now done: a full-range `mtest` over 3503 MiB passed with 0 errors, and `bdinfo` states the geometry independently as one `0xe0000000` bank. So the series can now say what was actually verified rather than implying it. Suggested wording for the commit message, since "Linux boots and reports the right size" was the weak part: > Verified on a Tronsmart Draco AW80 Telos with a full-range U-Boot mtest > (CONFIG_SYS_ALT_MEMTEST, address-line walk included) across 3503 MiB - everything below > U-Boot's own reservation - with zero errors. One caveat worth carrying: the top 81 MiB cannot be tested from U-Boot because U-Boot lives there. It is a small fraction and the address-line walk covers the full span either way, but it should be stated rather than glossed.