From: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
To: Saikiran <bjsaikiran@gmail.com>, jjohnson@kernel.org, kvalo@kernel.org
Cc: quic_bqiang@quicinc.com, linux-wireless@vger.kernel.org,
ath12k@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] wifi: ath12k: fix CMA error and MHI state mismatch during resume
Date: Tue, 3 Feb 2026 10:51:05 +0800 [thread overview]
Message-ID: <125f0ecb-79a5-4806-aa93-aecaf937885e@oss.qualcomm.com> (raw)
In-Reply-To: <20260202151720.49904-1-bjsaikiran@gmail.com>
On 2/2/2026 11:17 PM, Saikiran wrote:
> Commit 8d5f4da8d70b ("wifi: ath12k: support suspend/resume") introduced
> system suspend/resume support but caused a critical regression where
> CMA pages are corrupted during resume.
>
> 1. CMA page corruption:
> Calling mhi_unprepare_after_power_down() during suspend (via
> ATH12K_MHI_DEINIT) prematurely frees the fbc_image and rddm_image
> DMA buffers. When these pages are accessed during resume, the kernel
> detects corruption (Bad page state).
How, FBC image and RDDM image get re-allocated at resume, no?
>
> To fix this corruption, the driver must skip ATH12K_MHI_DEINIT during
> suspend, preserving the DMA buffers. However, implementing this fix
> exposes a second issue in the state machine:
>
> 2. Resume failure due to MHI state mismatch:
> When DEINIT is skipped during suspend to protect the memory, the
> ATH12K_MHI_INIT bit remains set. On resume, ath12k_mhi_start()
> blindly attempts to set INIT again, but the state machine rejects
> the transition:
>
> ath12k_wifi7_pci ...: failed to set mhi state INIT(0) in current
> mhi state (0x1)
>
> Fix the corruption and enable the correct suspend flow by:
>
> 1. In ath12k_mhi_stop(), skipping ATH12K_MHI_DEINIT if suspending.
> This prevents the memory corruption by keeping the device context
> valid (MHI_POWER_OFF_KEEP_DEV).
>
> 2. In ath12k_mhi_start(), checking if MHI_INIT is already set.
> This accommodates the new suspend flow where the device remains
> initialized, allowing the driver to proceed directly to POWER_ON.
>
> Tested with suspend/resume cycles on Qualcomm Snapdragon X Elite
> (SC8380XP) with WCN7850 WiFi. No CMA corruption observed, WiFi resumes
> successfully, and deep sleep works correctly.
>
> Fixes: 8d5f4da8d70b ("wifi: ath12k: support suspend/resume")
> Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.1.c5-00302 (Lenovo Yoga Slim 7x)
> Signed-off-by: Saikiran <bjsaikiran@gmail.com>
> ---
> drivers/net/wireless/ath/ath12k/mhi.c | 24 +++++++++++++++++-------
> 1 file changed, 17 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/net/wireless/ath/ath12k/mhi.c b/drivers/net/wireless/ath/ath12k/mhi.c
> index 45c0f66dcc5e..1a0b3bcc6bbf 100644
> --- a/drivers/net/wireless/ath/ath12k/mhi.c
> +++ b/drivers/net/wireless/ath/ath12k/mhi.c
> @@ -485,9 +485,14 @@ int ath12k_mhi_start(struct ath12k_pci *ab_pci)
>
> ab_pci->mhi_ctrl->timeout_ms = MHI_TIMEOUT_DEFAULT_MS;
>
> - ret = ath12k_mhi_set_state(ab_pci, ATH12K_MHI_INIT);
> - if (ret)
> - goto out;
> + /* In case of suspend/resume, MHI INIT is already done.
> + * So check if MHI INIT is set or not.
> + */
> + if (!test_bit(ATH12K_MHI_INIT, &ab_pci->mhi_state)) {
> + ret = ath12k_mhi_set_state(ab_pci, ATH12K_MHI_INIT);
> + if (ret)
> + goto out;
> + }
>
> ret = ath12k_mhi_set_state(ab_pci, ATH12K_MHI_POWER_ON);
> if (ret)
> @@ -501,16 +506,21 @@ int ath12k_mhi_start(struct ath12k_pci *ab_pci)
>
> void ath12k_mhi_stop(struct ath12k_pci *ab_pci, bool is_suspend)
> {
> - /* During suspend we need to use mhi_power_down_keep_dev()
> - * workaround, otherwise ath12k_core_resume() will timeout
> - * during resume.
> + /* During suspend, we need to use mhi_power_down_keep_dev()
> + * and avoid calling MHI_DEINIT. The deinit frees BHIE tables
> + * which causes memory corruption when those pages are
> + * accessed/freed again during resume. We want to keep the
> + * device prepared for resume, otherwise ath12k_core_resume()
> + * will timeout.
> */
> if (is_suspend)
> ath12k_mhi_set_state(ab_pci, ATH12K_MHI_POWER_OFF_KEEP_DEV);
> else
> ath12k_mhi_set_state(ab_pci, ATH12K_MHI_POWER_OFF);
>
> - ath12k_mhi_set_state(ab_pci, ATH12K_MHI_DEINIT);
> + /* Only deinit when doing full power down, not during suspend */
> + if (!is_suspend)
> + ath12k_mhi_set_state(ab_pci, ATH12K_MHI_DEINIT);
> }
>
> void ath12k_mhi_suspend(struct ath12k_pci *ab_pci)
next prev parent reply other threads:[~2026-02-03 2:51 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-02-02 15:17 Saikiran
2026-02-03 2:51 ` Baochen Qiang [this message]
2026-02-03 5:02 ` Jayasaikiran Banigallapati
2026-02-03 5:30 ` Baochen Qiang
2026-02-03 5:51 ` Jayasaikiran Banigallapati
2026-02-03 6:08 ` Baochen Qiang
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=125f0ecb-79a5-4806-aa93-aecaf937885e@oss.qualcomm.com \
--to=baochen.qiang@oss.qualcomm.com \
--cc=ath12k@lists.infradead.org \
--cc=bjsaikiran@gmail.com \
--cc=jjohnson@kernel.org \
--cc=kvalo@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-wireless@vger.kernel.org \
--cc=quic_bqiang@quicinc.com \
/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
all inboxes | Powered by JetHome®