[PATCH] spi: dln2: validate chip select count reported by the device

[PATCH] spi: dln2: validate chip select count reported by the device

[Farhad Alemi]

dln2_spi_get_cs_num() stores the chip select count reported by the device
straight into host->num_chipselect, and dln2_spi_cs_enable_all() then
evaluates GENMASK(num_chipselect - 1, 0), whose shift exponent is out of
bounds whenever that count is zero or larger than BITS_PER_LONG. The
resulting mask is handed back to the device in a u8 field, so any count
outside 1..8 is invalid regardless. Reject such counts with -EPROTO in
dln2_spi_get_cs_num(), where the value is first read.

Closes: https://lore.kernel.org/all/CA+0ovCioY634b646jDcs2cFmg_1q2UMN4wuTQ1V26Gy+Dm_c+A@mail.gmail.com/
Signed-off-by: Farhad Alemi <farhad.alemi@berkeley.edu>
---
The device was emulated.

--- a/drivers/spi/spi-dln2.c
+++ b/drivers/spi/spi-dln2.c
@@ -71,6 +71,8 @@
 #define DLN2_SPI_GET_MAX_DELAY_BETWEEN_FRAMES	DLN2_SPI_CMD(0x4C)

 #define DLN2_SPI_MAX_XFER_SIZE			256
+/* The CS bitmask is carried in a u8 on the wire, so 8 lines is the ceiling. */
+#define DLN2_SPI_MAX_CS				8
 #define DLN2_SPI_BUF_SIZE			(DLN2_SPI_MAX_XFER_SIZE + 16)
 #define DLN2_SPI_ATTR_LEAVE_SS_LOW		BIT(0)
 #define DLN2_TRANSFERS_WAIT_COMPLETE		1
@@ -201,6 +203,8 @@ static int dln2_spi_get_cs_num(struct dln2_spi
*dln2, u16 *cs_num)
 		return -EPROTO;

 	*cs_num = le16_to_cpu(rx.cs_count);
+	if (*cs_num == 0 || *cs_num > DLN2_SPI_MAX_CS)
+		return -EPROTO;

 	dev_dbg(&dln2->pdev->dev, "cs_num = %d\n", *cs_num);

Re: [PATCH] spi: dln2: validate chip select count reported by the device

[Andy Shevchenko]

On Thu, Sep 10, 2026 at 09:43:38PM +0000, Farhad Alemi wrote:
> dln2_spi_get_cs_num() stores the chip select count reported by the device
> straight into host->num_chipselect, and dln2_spi_cs_enable_all() then
> evaluates GENMASK(num_chipselect - 1, 0), whose shift exponent is out of
> bounds whenever that count is zero or larger than BITS_PER_LONG. The
> resulting mask is handed back to the device in a u8 field, so any count
> outside 1..8 is invalid regardless. Reject such counts with -EPROTO in
> dln2_spi_get_cs_num(), where the value is first read.

You need to check with the datasheet. It says that only 4 SS lines can be
programmed. The reserved bits must be held in 1. So, all in all, the _MAX_CS
should be 4.

> Closes: https://lore.kernel.org/all/CA+0ovCioY634b646jDcs2cFmg_1q2UMN4wuTQ1V26Gy+Dm_c+A@mail.gmail.com/
> Signed-off-by: Farhad Alemi <farhad.alemi@berkeley.edu>
> ---

Datasheet: https://diolan.com/media/wysiwyg/downloads/dln-api-manual.pdf

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] spi: dln2: validate chip select count reported by the device

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 352 bytes --]

On Thu, Sep 10, 2026 at 09:43:38PM +0000, Farhad Alemi wrote:
> dln2_spi_get_cs_num() stores the chip select count reported by the device
> straight into host->num_chipselect, and dln2_spi_cs_enable_all() then
> evaluates GENMASK(num_chipselect - 1, 0), whose shift exponent is out of

This doesn't apply against current code, please check and resend.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]