| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <96e9898f-e306-4b73-9d06-5515916cd769@nvidia.com>
Date: Fri, 24 Oct 2025 16:42:06 +0100
From: Jon Hunter <jonathanh@...dia.com>
To: Kartik Rajput <kkartik@...dia.com>, akhilrajeev@...dia.com,
andi.shyti@...nel.org, robh@...nel.org, krzk+dt@...nel.org,
conor+dt@...nel.org, thierry.reding@...il.com, ldewangan@...dia.com,
digetx@...il.com, linux-i2c@...r.kernel.org, devicetree@...r.kernel.org,
linux-tegra@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v9 3/4] i2c: tegra: Add support for SW mutex register
On 01/10/2025 07:47, Kartik Rajput wrote:
> Add support for SW mutex register introduced in Tegra264 to provide
> an option to share the interface between multiple firmwares and/or
> VMs. This involves following steps:
>
> - A firmware/OS writes its unique ID to the mutex REQUEST field.
> - Ownership is established when reading the GRANT field returns the
> same ID.
> - If GRANT shows a different non-zero ID, the firmware/OS retries
> until timeout.
> - After completing access, it releases the mutex by writing 0.
>
> However, the hardware does not ensure any protection based on the
> values. The driver/firmware should honor the peer who already holds
> the mutex.
>
> Signed-off-by: Kartik Rajput <kkartik@...dia.com>
> Signed-off-by: Akhil R <akhilrajeev@...dia.com>
> ---
> v7 -> v8:
> * Use `bool` instead of `int` for `locked` variable in
> tegra_i2c_mutex_lock() function.
> v6 -> v7:
> * Return bool from tegra_i2c_mutex_acquired() and
> tegra_i2c_mutex_trylock() functions.
> * Move `has_mutex` check inside tegra_i2c_mutex_lock/unlock
> functions.
> * Remove redundant empty line added in tegra_i2c_xfer() in v6.
> * Fix pm_runtime_put() not getting called if mutex unlock fails.
> * In tegra_i2c_mutex_lock() simplify the logic to check if the
> mutex is acquired or not by checking the value of `ret`
> variable.
> * Update commit message to describe the functioning of SW mutex
> feature.
> v4 -> v6:
> * Guard tegra_i2c_mutex_lock() and tegra_i2c_mutex_unlock() to
> ensure that they are called on platforms which support SW
> mutex.
> v3 -> v4:
> * Update timeout logic of tegra_i2c_mutex_lock() to use
> read_poll_timeout APIs for improving timeout logic.
> * Add tegra_i2c_mutex_acquired() to check if mutex is acquired
> or not.
> * Rename I2C_SW_MUTEX_ID as I2C_SW_MUTEX_ID_CCPLEX.
> * Function tegra_i2c_poll_register() was moved unnecessarily, it
> has now been moved to its original location.
> * Use tegra_i2c_mutex_lock/unlock APIs in the tegra_i2c_xfer()
> function. This ensures proper propagation of error in case
> mutex lock fails.
> Please note that as the function tegra_i2c_xfer() is
> already guarded by the bus lock operation there is no need of
> additional lock for the tegra_i2c_mutex_lock/unlock APIs.
> v2 -> v3:
> * Update tegra_i2c_mutex_trylock and tegra_i2c_mutex_unlock to
> use readl and writel APIs instead of i2c_readl and i2c_writel
> which use relaxed APIs.
> * Use dev_warn instead of WARN_ON if mutex lock/unlock fails.
> v1 -> v2:
> * Fixed typos.
> * Fix tegra_i2c_mutex_lock() logic.
> * Add a timeout in tegra_i2c_mutex_lock() instead of polling for
> mutex indefinitely.
> ---
> drivers/i2c/busses/i2c-tegra.c | 92 ++++++++++++++++++++++++++++++++++
> 1 file changed, 92 insertions(+)
>
> diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c
> index cc75340f6cb5..1c8c24ae54ed 100644
> --- a/drivers/i2c/busses/i2c-tegra.c
> +++ b/drivers/i2c/busses/i2c-tegra.c
> @@ -137,6 +137,14 @@
>
> #define I2C_MASTER_RESET_CNTRL 0x0a8
>
> +#define I2C_SW_MUTEX 0x0ec
> +#define I2C_SW_MUTEX_REQUEST GENMASK(3, 0)
> +#define I2C_SW_MUTEX_GRANT GENMASK(7, 4)
> +#define I2C_SW_MUTEX_ID_CCPLEX 9
> +
> +/* SW mutex acquire timeout value in microseconds. */
> +#define I2C_SW_MUTEX_TIMEOUT_US (25 * USEC_PER_MSEC)
> +
> /* configuration load timeout in microseconds */
> #define I2C_CONFIG_LOAD_TIMEOUT 1000000
>
> @@ -210,6 +218,7 @@ enum msg_end_type {
> * @has_interface_timing_reg: Has interface timing register to program the tuned
> * timing settings.
> * @has_hs_mode_support: Has support for high speed (HS) mode transfers.
> + * @has_mutex: Has mutex register for mutual exclusion with other firmwares or VMs.
> */
> struct tegra_i2c_hw_feature {
> bool has_continue_xfer_support;
> @@ -237,6 +246,7 @@ struct tegra_i2c_hw_feature {
> u32 setup_hold_time_hs_mode;
> bool has_interface_timing_reg;
> bool has_hs_mode_support;
> + bool has_mutex;
> };
>
> /**
> @@ -381,6 +391,76 @@ static void i2c_readsl(struct tegra_i2c_dev *i2c_dev, void *data,
> readsl(i2c_dev->base + tegra_i2c_reg_addr(i2c_dev, reg), data, len);
> }
>
> +static bool tegra_i2c_mutex_acquired(struct tegra_i2c_dev *i2c_dev)
> +{
> + unsigned int reg = tegra_i2c_reg_addr(i2c_dev, I2C_SW_MUTEX);
> + u32 val, id;
> +
> + val = readl(i2c_dev->base + reg);
> + id = FIELD_GET(I2C_SW_MUTEX_GRANT, val);
> +
> + return id == I2C_SW_MUTEX_ID_CCPLEX;
> +}
> +
> +static bool tegra_i2c_mutex_trylock(struct tegra_i2c_dev *i2c_dev)
> +{
> + unsigned int reg = tegra_i2c_reg_addr(i2c_dev, I2C_SW_MUTEX);
> + u32 val, id;
> +
> + val = readl(i2c_dev->base + reg);
> + id = FIELD_GET(I2C_SW_MUTEX_GRANT, val);
> + if (id != 0 && id != I2C_SW_MUTEX_ID_CCPLEX)
> + return false;
> +
> + val = FIELD_PREP(I2C_SW_MUTEX_REQUEST, I2C_SW_MUTEX_ID_CCPLEX);
> + writel(val, i2c_dev->base + reg);
> +
> + return tegra_i2c_mutex_acquired(i2c_dev);
> +}
> +
> +static int tegra_i2c_mutex_lock(struct tegra_i2c_dev *i2c_dev)
> +{
> + bool locked;
> + int ret;
> +
> + if (!i2c_dev->hw->has_mutex)
> + return 0;
> +
> + if (i2c_dev->atomic_mode)
> + ret = read_poll_timeout_atomic(tegra_i2c_mutex_trylock, locked, locked,
> + USEC_PER_MSEC, I2C_SW_MUTEX_TIMEOUT_US,
> + false, i2c_dev);
> + else
> + ret = read_poll_timeout(tegra_i2c_mutex_trylock, locked, locked, USEC_PER_MSEC,
> + I2C_SW_MUTEX_TIMEOUT_US, false, i2c_dev);
> +
> + if (ret)
> + dev_warn(i2c_dev->dev, "failed to acquire mutex\n");
> +
> + return ret;
> +}
> +
> +static int tegra_i2c_mutex_unlock(struct tegra_i2c_dev *i2c_dev)
> +{
> + unsigned int reg = tegra_i2c_reg_addr(i2c_dev, I2C_SW_MUTEX);
> + u32 val, id;
> +
> + if (!i2c_dev->hw->has_mutex)
> + return 0;
> +
> + val = readl(i2c_dev->base + reg);
> +
> + id = FIELD_GET(I2C_SW_MUTEX_GRANT, val);
> + if (id && id != I2C_SW_MUTEX_ID_CCPLEX) {
> + dev_warn(i2c_dev->dev, "unable to unlock mutex, mutex is owned by: %u\n", id);
> + return -EPERM;
> + }
> +
> + writel(0, i2c_dev->base + reg);
> +
> + return 0;
> +}
> +
> static void tegra_i2c_mask_irq(struct tegra_i2c_dev *i2c_dev, u32 mask)
> {
> u32 int_mask;
> @@ -1432,6 +1512,10 @@ static int tegra_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg msgs[],
> return ret;
> }
>
> + ret = tegra_i2c_mutex_lock(i2c_dev);
> + if (ret)
> + return ret;
> +
I wonder if it would be better to have a wrapper function around
tegra_i2c_xfer() called tegra264_i2c_xfer() that is only used for
Tegra264 platforms and invokes these sw-mutex functions?
Jon
--
nvpublic
Powered by blists - more mailing lists