Linux Input/HID development
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Emfox Zhou" <emfox@debian.org>
Cc: linux-input@vger.kernel.org, dmitry.torokhov@gmail.com
Subject: Re: [PATCH v4] HID: intel-thc-hid: intel-thc: restore PRD table capacity
Date: Tue, 17 Nov 2026 00:51:49 +0000	[thread overview]
Message-ID: <sashiko-outbox-218028@kernel.org> (raw)
In-Reply-To: <20261117003247.940876-1-emfox@debian.org>

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [High] HID: intel-thc-hid: race between thc_dma_read() and DMA pointer
  reset
--

commit 16c1e87f8c9f9a7660927e9a01abdc216355c96d
Author: Emfox Zhou <emfox@debian.org>

HID: intel-thc-hid: intel-thc: restore PRD table capacity

This commit rebuilds the RxDMA and SWDMA PRD tables before starting the DMA
engines to restore entry capacities overwritten by unread frames, and adds a
mutex to serialize table rebuilds with thc_dma_read().

Link: https://lore.kernel.org/r/20261115040756.602092-1-emfox@debian.org

> diff --git a/drivers/hid/intel-thc-hid/intel-thc/intel-thc-dma.c b/drivers/hid/intel-thc-hid/intel-thc/intel-thc-dma.c
> index c80ef8c205742..413f264e30106 100644
> --- a/drivers/hid/intel-thc-hid/intel-thc/intel-thc-dma.c
> +++ b/drivers/hid/intel-thc-hid/intel-thc/intel-thc-dma.c
> @@ -211,12 +211,27 @@ static void thc_copy_sgls_to_prd(struct thc_device *dev,
>  {
>  	unsigned int i;
>  
> +	guard(mutex)(&dev->dma_ctx->prd_lock);
> +
> +	if (!config->is_enabled || !config->prd_tbls)
> +		return;
> +
>  	memset(config->prd_tbls, 0, array_size(PRD_TABLE_SIZE, config->prd_tbl_num));
>  
>  	for (i = 0; i < config->prd_tbl_num; i++)
>  		thc_copy_one_sgl_to_prd(dev, config, i);
>  }
>  
> +/*
> + * A frame that is never read leaves its entry short, and the engine reads
> + * the table again on every start: put the capacity back.
> + */
> +static void thc_rebuild_rx_prd(struct thc_device *dev)
> +{
> +	thc_copy_sgls_to_prd(dev, &dev->dma_ctx->dma_config[THC_RXDMA1]);
> +	thc_copy_sgls_to_prd(dev, &dev->dma_ctx->dma_config[THC_RXDMA2]);
> +}

[ ... ]

> @@ -491,6 +507,9 @@ int thc_dma_configure(struct thc_device *dev)
>  		return -EINVAL;
>  	}
>  
> +	/* An entry may still be short from a frame nobody read. */
> +	thc_rebuild_rx_prd(dev);
> +
>  	for (chan = 0; chan < MAX_THC_DMA_CHANNEL; chan++) {
>  		dma_set_prd_base_addr(dev,
>  				      dma_ctx->dma_config[chan].prd_tbls_dma_handle,
> @@ -685,6 +704,8 @@ static int thc_dma_read(struct thc_device *dev,
>  	u8 read_ptr, write_ptr, prd_table_index;
>  	int status;
>  
> +	guard(mutex)(&dev->dma_ctx->prd_lock);
> +
>  	if (!is_dma_buf_empty(dev, read_config, &read_ptr, &write_ptr)) {
>  		prd_table_index = write_ptr & THC_POINTER_MASK;
>  
> @@ -765,11 +786,16 @@ static int thc_swdma_read_start(struct thc_device *dev, void *write_buff,
>  	if (ret)
>  		return ret;
>  
> -	if (thc_wait_for_dma_pause(dev, THC_RXDMA1) || thc_wait_for_dma_pause(dev, THC_RXDMA2))
> +	if (thc_wait_for_dma_pause(dev, THC_RXDMA1) ||
> +	    thc_wait_for_dma_pause(dev, THC_RXDMA2) ||
> +	    thc_wait_for_dma_pause(dev, THC_SWDMA))
>  		return -EIO;
>  
>  	thc_reset_dma_settings(dev);
>  
> +	/* A transfer that timed out left its frame length in the entry. */
> +	thc_copy_sgls_to_prd(dev, &dev->dma_ctx->dma_config[THC_SWDMA]);
> +
>  	/*
>  	 * Max input size control feature is only available for RxDMA, it must keep disabled
>  	 * during SWDMA operation, and restore to previous state after SWDMA is done.
> @@ -854,6 +880,9 @@ static int thc_swdma_read_completion(struct thc_device *dev)
>  
>  	thc_reset_dma_settings(dev);
>  
> +	/* The receive engine starts again just below. */
> +	thc_rebuild_rx_prd(dev);
> +
>  	dma_set_start_bit(dev, &dev->dma_ctx->dma_config[THC_RXDMA2]);
>  
>  	ret = thc_interrupt_quiesce(dev, false);

[Severity: High]
Can thc_dma_read() still race with thc_reset_dma_settings() and
dma_set_start_bit() because prd_lock only covers thc_copy_sgls_to_prd()?

is_dma_buf_empty() treats the ring as empty only when the masked TPCRP and
TPCWP pointers match with different wraparound bits.  thc_reset_dma_settings()
rewinds TPCRP to 0 without setting TPCWP to THC_POINTER_WRAPAROUND (0x80), and
TPCWP is only set to 0x80 when dma_set_start_bit() restarts THC_RXDMA2.
Because both register updates happen outside prd_lock in thc_dma_configure(),
thc_swdma_read_start() and thc_swdma_read_completion(), the threaded IRQ
handler can enter thc_dma_read() in between:

CPU0                                  CPU1 (threaded IRQ)
thc_swdma_read_completion()
  thc_reset_dma_settings()
    /* TPCRP = 0, TPCWP != 0x80 */
  thc_rebuild_rx_prd()
    thc_copy_sgls_to_prd()
      lock(&dma_ctx->prd_lock)
      /* restores entry.len = SZ_4K */
      unlock(&dma_ctx->prd_lock)
                                      quickspi_irq_thread_handler()
                                        thc_rxdma_read()
                                          thc_dma_read()
                                            lock(&dma_ctx->prd_lock)
                                            is_dma_buf_empty() -> false
                                            read_dma_buffer()  <-- buffer overflow
                                            update_write_pointer()
                                            unlock(&dma_ctx->prd_lock)
  dma_set_start_bit()

When thc_dma_read() sees the non-empty pointers after thc_rebuild_rx_prd(),
read_dma_buffer() copies a rebuilt PRD table entry whose length was restored
to max_packet_size (aligned to SZ_4K) instead of a frame length written by
hardware, overflowing qsdev->input_buf in intel-quickspi when
HIDSPI_INPUT_BODY_SIZE(max_input_len) is smaller than SZ_4K and corrupting
TPCWP.  The same window remains open across the entire SWDMA transfer after
thc_swdma_read_start() calls thc_reset_dma_settings() while THC_RXDMA2 is
stopped.

Even if the reset and dma_set_start_bit() complete before the woken IRQ thread
calls thc_dma_read(), is_dma_buf_empty() now returns true on the first call,
so thc_dma_read() returns 0 without setting *read_len:

drivers/hid/intel-thc-hid/intel-quickspi/pci-quickspi.c:quickspi_irq_thread_handler() {
	...
		while (!read_finished) {
			ret = thc_rxdma_read(qsdev->thc_hw, THC_RXDMA2, qsdev->input_buf,
					     &input_len, &read_finished);
			if (ret) {
				err_recover = 1;
				goto end;
			}

			quickspi_handle_input_data(qsdev, input_len);
		}
}

quickspi_irq_thread_handler() then passes an uninitialized input_len and stale
qsdev->input_buf to quickspi_handle_input_data() (and quicki2c's
handle_input_report() replays stale data in qcdev->input_buf).

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117003247.940876-1-emfox@debian.org?part=1

  reply	other threads:[~2026-11-17  0:51 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17  0:32 [PATCH v4] HID: intel-thc-hid: intel-thc: restore PRD table capacity Emfox Zhou
2026-11-17  0:51 ` sashiko-bot [this message]
  -- strict thread matches above, loose matches on Subject: below --
2026-11-17  0:27 Emfox Zhou
2026-11-17  0:48 ` sashiko-bot

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-218028@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=dmitry.torokhov@gmail.com \
    --cc=emfox@debian.org \
    --cc=linux-input@vger.kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox