Change the scanning /memreserve/ and /reserved-memory node order to fix
Kernel panic on Khadas Vim3 Board.
If /memreserve/ goes first, the memory is reserved, but nomap can't be
applied to the region. So the memory won't be used by Linux, but it is
still present in the linear map as normal memory, which allows
speculation. Legitimate access to adjacent pages will cause the CPU
to end up prefetching into them leading to Kernel panic.
So /reserved-memory node should go first, as it has a more updated
description of the memory regions and can apply flags, like nomap.
Signed-off-by: Lucas Tanure <[email protected]>i
---
Previous Threads:
#regzbot link: https://lore.kernel.org/linux-arm-kernel/[email protected]/#regzbot
#regzbot link: https://lore.kernel.org/all/CAJX_Q+1Tjc+-TjZ6JW9X0NxEdFe=82a9626yL63j7uVD4LpxEA@mail.gmail.com/
V1: https://lore.kernel.org/all/[email protected]/
v2: https://lore.kernel.org/all/[email protected]/
v3: https://lore.kernel.org/linux-mm/[email protected]/T/
Change from V3:
- Drop cover letter
- Add more information into the patch
Change from V2:
- Remove region overlap and reserved checks
Change from V1:
- Instead of allowing to mark nomap a region already reserved, give
precedence to /reserved-memory node scan.
---
drivers/of/fdt.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
index d14735a81301..bf502ba8da95 100644
--- a/drivers/of/fdt.c
+++ b/drivers/of/fdt.c
@@ -635,6 +635,9 @@ void __init early_init_fdt_scan_reserved_mem(void)
if (!initial_boot_params)
return;
+ fdt_scan_reserved_mem();
+ fdt_reserve_elfcorehdr();
+
/* Process header /memreserve/ fields */
for (n = 0; ; n++) {
fdt_get_mem_rsv(initial_boot_params, n, &base, &size);
@@ -643,8 +646,6 @@ void __init early_init_fdt_scan_reserved_mem(void)
memblock_reserve(base, size);
}
- fdt_scan_reserved_mem();
- fdt_reserve_elfcorehdr();
fdt_init_reserved_mem();
}
--
2.40.0
On Mon, 24 Apr 2023 12:38:46 +0100, Lucas Tanure wrote:
> Change the scanning /memreserve/ and /reserved-memory node order to fix
> Kernel panic on Khadas Vim3 Board.
>
> If /memreserve/ goes first, the memory is reserved, but nomap can't be
> applied to the region. So the memory won't be used by Linux, but it is
> still present in the linear map as normal memory, which allows
> speculation. Legitimate access to adjacent pages will cause the CPU
> to end up prefetching into them leading to Kernel panic.
>
> So /reserved-memory node should go first, as it has a more updated
> description of the memory regions and can apply flags, like nomap.
>
> Signed-off-by: Lucas Tanure <[email protected]>i
> ---
> Previous Threads:
> #regzbot link: https://lore.kernel.org/linux-arm-kernel/[email protected]/#regzbot
> #regzbot link: https://lore.kernel.org/all/CAJX_Q+1Tjc+-TjZ6JW9X0NxEdFe=82a9626yL63j7uVD4LpxEA@mail.gmail.com/
Note that this should have been a 'Link:' entry in the commit message
for regzbot. I added it.
>
> V1: https://lore.kernel.org/all/[email protected]/
> v2: https://lore.kernel.org/all/[email protected]/
> v3: https://lore.kernel.org/linux-mm/[email protected]/T/
>
> Change from V3:
> - Drop cover letter
> - Add more information into the patch
>
> Change from V2:
> - Remove region overlap and reserved checks
>
> Change from V1:
> - Instead of allowing to mark nomap a region already reserved, give
> precedence to /reserved-memory node scan.
> ---
> drivers/of/fdt.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
Applied, thanks!
Going to let this sit in linux-next for a bit. I also didn't tag for
stable kernels. I want it to get some exposure in mainline at least for
a few -rc releases rather than getting backported right away. Please
help ensure that happens if you see stable review emails. I'm going to
be offline most of the next month.
Rob