The Reed-Solomon library enforces the constraint `n <= 2^m - 1` via a
BUG_ON() [1], where `n` is `block_size + ecc_size` and `m` is `symsize`
for the pstore RAM backend. A driver providing invalid parameters can
trigger this, leading to a kernel panic. For more details on the theory
behind, see [2].

This issue was discovered during developing chromeos_pstore driver.
Link: https://lore.kernel.org/lkml/20250610050458.4014083-1-naoyatezuka@chromium.org/

Add a check to validate this constraint before initializing Reed-Solomon
codec. On failure, return -EINVAL to prevent the panic.

[1] https://elixir.bootlin.com/linux/v6.15/source/lib/reed_solomon/decode_rs.c#L43
[2] https://www.cs.cmu.edu/~guyb/realworld/reedsolomon/reed_solomon_codes.html

Signed-off-by: Naoya Tezuka <naoyatezuka@chromium.org>
---
 fs/pstore/ram_core.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/fs/pstore/ram_core.c b/fs/pstore/ram_core.c
index f1848cdd6d34..c7a2ff9c5a6c 100644
--- a/fs/pstore/ram_core.c
+++ b/fs/pstore/ram_core.c
@@ -212,6 +212,14 @@ static int persistent_ram_init_ecc(struct persistent_ram_zone *prz,
 		return -EINVAL;
 	}
 
+	if (prz->ecc_info.block_size + prz->ecc_info.ecc_size >
+	    (1 << prz->ecc_info.symsize) - 1) {
+		pr_err("%s: invalid ecc parameters (block_size = %d, ecc_size = %d, symsize = %d\n",
+		       __func__, prz->ecc_info.block_size,
+		       prz->ecc_info.ecc_size, prz->ecc_info.symsize);
+		return -EINVAL;
+	}
+
 	prz->buffer_size -= ecc_total;
 	prz->par_buffer = buffer->data + prz->buffer_size;
 	prz->par_header = prz->par_buffer +
-- 
2.50.0.rc2.701.gf1e915cc24-goog

On Mon, Jun 23, 2025 at 03:28:27PM +0900, Naoya Tezuka wrote:
> The Reed-Solomon library enforces the constraint `n <= 2^m - 1` via a
> BUG_ON() [1], where `n` is `block_size + ecc_size` and `m` is `symsize`
> for the pstore RAM backend. A driver providing invalid parameters can
> trigger this, leading to a kernel panic. For more details on the theory
> behind, see [2].
> 
> This issue was discovered during developing chromeos_pstore driver.
> Link: https://lore.kernel.org/lkml/20250610050458.4014083-1-naoyatezuka@chromium.org/

I'd prefer to unify it by using a [3] or at least move the tag to the end of
commit message.

> Add a check to validate this constraint before initializing Reed-Solomon
> codec. On failure, return -EINVAL to prevent the panic.
> 
> [1] https://elixir.bootlin.com/linux/v6.15/source/lib/reed_solomon/decode_rs.c#L43
> [2] https://www.cs.cmu.edu/~guyb/realworld/reedsolomon/reed_solomon_codes.html
> 
> Signed-off-by: Naoya Tezuka <naoyatezuka@chromium.org>

It should preserve my R-b tag as v2 doesn't change too much. Anyway,
Reviewed-by: Tzung-Bi Shih <tzungbi@kernel.org>