diff mbox series

[3/4] coresight: tmc-etr: Fix race condition between sysfs and perf mode

Message ID 20241202092419.11777-4-yangyicong@huawei.com (mailing list archive)
State New
Headers show
Series Coresight TMC-ETR some bugfixes and cleanups | expand

Commit Message

Yicong Yang Dec. 2, 2024, 9:24 a.m. UTC
From: Yicong Yang <yangyicong@hisilicon.com>

When trying to run perf and sysfs mode simultaneously, the WARN_ON()
in tmc_etr_enable_hw() is triggered sometimes:

 WARNING: CPU: 42 PID: 3911571 at drivers/hwtracing/coresight/coresight-tmc-etr.c:1060 tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc]
 [..snip..]
 Call trace:
  tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc] (P)
  tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc] (L)
  tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc]
  coresight_enable_path+0x1c8/0x218 [coresight]
  coresight_enable_sysfs+0xa4/0x228 [coresight]
  enable_source_store+0x58/0xa8 [coresight]
  dev_attr_store+0x20/0x40
  sysfs_kf_write+0x4c/0x68
  kernfs_fop_write_iter+0x120/0x1b8
  vfs_write+0x2c8/0x388
  ksys_write+0x74/0x108
  __arm64_sys_write+0x24/0x38
  el0_svc_common.constprop.0+0x64/0x148
  do_el0_svc+0x24/0x38
  el0_svc+0x3c/0x130
  el0t_64_sync_handler+0xc8/0xd0
  el0t_64_sync+0x1ac/0x1b0
 ---[ end trace 0000000000000000 ]---

Since the enablement of sysfs mode is separeted into two critical regions,
one for sysfs buffer allocation and another for hardware enablement, it's
possible to race with the perf mode. Fix this by double check whether
the perf mode's been used before enabling the hardware in sysfs mode.

Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 .../hwtracing/coresight/coresight-tmc-etr.c   | 30 +++++++++++++++++++
 1 file changed, 30 insertions(+)

Comments

Suzuki K Poulose Dec. 2, 2024, 9:46 a.m. UTC | #1
Hi Yicong

On 02/12/2024 09:24, Yicong Yang wrote:
> From: Yicong Yang <yangyicong@hisilicon.com>
> 
> When trying to run perf and sysfs mode simultaneously, the WARN_ON()
> in tmc_etr_enable_hw() is triggered sometimes:
> 
>   WARNING: CPU: 42 PID: 3911571 at drivers/hwtracing/coresight/coresight-tmc-etr.c:1060 tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc]
>   [..snip..]
>   Call trace:
>    tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc] (P)
>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc] (L)
>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc]
>    coresight_enable_path+0x1c8/0x218 [coresight]
>    coresight_enable_sysfs+0xa4/0x228 [coresight]
>    enable_source_store+0x58/0xa8 [coresight]
>    dev_attr_store+0x20/0x40
>    sysfs_kf_write+0x4c/0x68
>    kernfs_fop_write_iter+0x120/0x1b8
>    vfs_write+0x2c8/0x388
>    ksys_write+0x74/0x108
>    __arm64_sys_write+0x24/0x38
>    el0_svc_common.constprop.0+0x64/0x148
>    do_el0_svc+0x24/0x38
>    el0_svc+0x3c/0x130
>    el0t_64_sync_handler+0xc8/0xd0
>    el0t_64_sync+0x1ac/0x1b0
>   ---[ end trace 0000000000000000 ]---
> 
> Since the enablement of sysfs mode is separeted into two critical regions,
> one for sysfs buffer allocation and another for hardware enablement, it's
> possible to race with the perf mode. Fix this by double check whether
> the perf mode's been used before enabling the hardware in sysfs mode.

Thanks for the fix. Some minor comments below.

It needs a Fixes tag.

> 
> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
> ---
>   .../hwtracing/coresight/coresight-tmc-etr.c   | 30 +++++++++++++++++++
>   1 file changed, 30 insertions(+)
> 
> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/drivers/hwtracing/coresight/coresight-tmc-etr.c
> index ad83714ca4dc..d382d95da5ff 100644
> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
> @@ -1230,6 +1230,36 @@ static int tmc_enable_etr_sink_sysfs(struct coresight_device *csdev)
>   
>   	spin_lock_irqsave(&drvdata->spinlock, flags);
>   
> +	/*
> +	 * Since the sysfs buffer allocation and the hardware enablement is not
> +	 * in the same critical region, it's possible to race with the perf
> +	 * mode:
> +	 *   [sysfs mode]                   [perf mode]
> +	 *   tmc_etr_get_sysfs_buffer()
> +	 *     spin_lock(&drvdata->spinlock)
> +	 *     [sysfs buffer allocation]
> +	 *     spin_unlock(&drvdata->spinlock)
> +	 *                                  spin_lock(&drvdata->spinlock)
> +	 *                                  tmc_etr_enable_hw()
> +	 *                                    drvdata->etr_buf = etr_perf->etr_buf
> +	 *                                  spin_unlock(&drvdata->spinlock)
> +	 *   spin_lock(&drvdata->spinlock)
> +	 *   tmc_etr_enable_hw()
> +	 *     WARN_ON(drvdata->etr_buf) // WARN sicne etr_buf initialized at
> +	 *                                  the perf side
> +	 *   spin_unlock(&drvdata->spinlock)
> +	 *
> +	 * So check here before continue.
> +	 */
> +	if (coresight_get_mode(csdev) == CS_MODE_PERF) {
> +		drvdata->sysfs_buf = NULL;
> +		spin_unlock_irqrestore(&drvdata->spinlock, flags);
> +
> +		/* Free allocated memory out side of the spinlock */
> +		tmc_etr_free_sysfs_buf(sysfs_buf);
> +		return -EBUSY;
> +	}

With this in place, I think we should remove the check for CS_MODE_PERF 
in get_etr_sysfs_buf() to avoid confusion (which I believe opened up 
this race)

Suzuki



> +
>   	/*
>   	 * In sysFS mode we can have multiple writers per sink.  Since this
>   	 * sink is already enabled no memory is needed and the HW need not be
Yicong Yang Dec. 2, 2024, 10:54 a.m. UTC | #2
On 2024/12/2 17:46, Suzuki K Poulose wrote:
> Hi Yicong
> 
> On 02/12/2024 09:24, Yicong Yang wrote:
>> From: Yicong Yang <yangyicong@hisilicon.com>
>>
>> When trying to run perf and sysfs mode simultaneously, the WARN_ON()
>> in tmc_etr_enable_hw() is triggered sometimes:
>>
>>   WARNING: CPU: 42 PID: 3911571 at drivers/hwtracing/coresight/coresight-tmc-etr.c:1060 tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc]
>>   [..snip..]
>>   Call trace:
>>    tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc] (P)
>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc] (L)
>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc]
>>    coresight_enable_path+0x1c8/0x218 [coresight]
>>    coresight_enable_sysfs+0xa4/0x228 [coresight]
>>    enable_source_store+0x58/0xa8 [coresight]
>>    dev_attr_store+0x20/0x40
>>    sysfs_kf_write+0x4c/0x68
>>    kernfs_fop_write_iter+0x120/0x1b8
>>    vfs_write+0x2c8/0x388
>>    ksys_write+0x74/0x108
>>    __arm64_sys_write+0x24/0x38
>>    el0_svc_common.constprop.0+0x64/0x148
>>    do_el0_svc+0x24/0x38
>>    el0_svc+0x3c/0x130
>>    el0t_64_sync_handler+0xc8/0xd0
>>    el0t_64_sync+0x1ac/0x1b0
>>   ---[ end trace 0000000000000000 ]---
>>
>> Since the enablement of sysfs mode is separeted into two critical regions,
>> one for sysfs buffer allocation and another for hardware enablement, it's
>> possible to race with the perf mode. Fix this by double check whether
>> the perf mode's been used before enabling the hardware in sysfs mode.
> 
> Thanks for the fix. Some minor comments below.
> 
> It needs a Fixes tag.
> 
>>
>> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
>> ---
>>   .../hwtracing/coresight/coresight-tmc-etr.c   | 30 +++++++++++++++++++
>>   1 file changed, 30 insertions(+)
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> index ad83714ca4dc..d382d95da5ff 100644
>> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> @@ -1230,6 +1230,36 @@ static int tmc_enable_etr_sink_sysfs(struct coresight_device *csdev)
>>         spin_lock_irqsave(&drvdata->spinlock, flags);
>>   +    /*
>> +     * Since the sysfs buffer allocation and the hardware enablement is not
>> +     * in the same critical region, it's possible to race with the perf
>> +     * mode:
>> +     *   [sysfs mode]                   [perf mode]
>> +     *   tmc_etr_get_sysfs_buffer()
>> +     *     spin_lock(&drvdata->spinlock)
>> +     *     [sysfs buffer allocation]
>> +     *     spin_unlock(&drvdata->spinlock)
>> +     *                                  spin_lock(&drvdata->spinlock)
>> +     *                                  tmc_etr_enable_hw()
>> +     *                                    drvdata->etr_buf = etr_perf->etr_buf
>> +     *                                  spin_unlock(&drvdata->spinlock)
>> +     *   spin_lock(&drvdata->spinlock)
>> +     *   tmc_etr_enable_hw()
>> +     *     WARN_ON(drvdata->etr_buf) // WARN sicne etr_buf initialized at
>> +     *                                  the perf side
>> +     *   spin_unlock(&drvdata->spinlock)
>> +     *
>> +     * So check here before continue.
>> +     */
>> +    if (coresight_get_mode(csdev) == CS_MODE_PERF) {
>> +        drvdata->sysfs_buf = NULL;
>> +        spin_unlock_irqrestore(&drvdata->spinlock, flags);
>> +
>> +        /* Free allocated memory out side of the spinlock */
>> +        tmc_etr_free_sysfs_buf(sysfs_buf);
>> +        return -EBUSY;
>> +    }
> 
> With this in place, I think we should remove the check for CS_MODE_PERF in get_etr_sysfs_buf() to avoid confusion (which I believe opened up this race)
> 

ok, considering that maybe we can optimize it a bit more in the tmc_etr_get_sysfs_buffer().
Check whether the perf mode's already running before we actually allocate the buffer.
Then we can save the time of allocating/freeing the sysfs buffer if race with the perf mode.

Thanks.
James Clark Dec. 2, 2024, 11:19 a.m. UTC | #3
On 02/12/2024 9:46 am, Suzuki K Poulose wrote:
> Hi Yicong
> 
> On 02/12/2024 09:24, Yicong Yang wrote:
>> From: Yicong Yang <yangyicong@hisilicon.com>
>>
>> When trying to run perf and sysfs mode simultaneously, the WARN_ON()
>> in tmc_etr_enable_hw() is triggered sometimes:
>>
>>   WARNING: CPU: 42 PID: 3911571 at drivers/hwtracing/coresight/ 
>> coresight-tmc-etr.c:1060 tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc]
>>   [..snip..]
>>   Call trace:
>>    tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc] (P)
>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc] (L)
>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc]
>>    coresight_enable_path+0x1c8/0x218 [coresight]
>>    coresight_enable_sysfs+0xa4/0x228 [coresight]
>>    enable_source_store+0x58/0xa8 [coresight]
>>    dev_attr_store+0x20/0x40
>>    sysfs_kf_write+0x4c/0x68
>>    kernfs_fop_write_iter+0x120/0x1b8
>>    vfs_write+0x2c8/0x388
>>    ksys_write+0x74/0x108
>>    __arm64_sys_write+0x24/0x38
>>    el0_svc_common.constprop.0+0x64/0x148
>>    do_el0_svc+0x24/0x38
>>    el0_svc+0x3c/0x130
>>    el0t_64_sync_handler+0xc8/0xd0
>>    el0t_64_sync+0x1ac/0x1b0
>>   ---[ end trace 0000000000000000 ]---
>>
>> Since the enablement of sysfs mode is separeted into two critical 
>> regions,
>> one for sysfs buffer allocation and another for hardware enablement, it's
>> possible to race with the perf mode. Fix this by double check whether
>> the perf mode's been used before enabling the hardware in sysfs mode.
> 
> Thanks for the fix. Some minor comments below.
> 
> It needs a Fixes tag.
> 
>>
>> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
>> ---
>>   .../hwtracing/coresight/coresight-tmc-etr.c   | 30 +++++++++++++++++++
>>   1 file changed, 30 insertions(+)
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/ 
>> drivers/hwtracing/coresight/coresight-tmc-etr.c
>> index ad83714ca4dc..d382d95da5ff 100644
>> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> @@ -1230,6 +1230,36 @@ static int tmc_enable_etr_sink_sysfs(struct 
>> coresight_device *csdev)
>>       spin_lock_irqsave(&drvdata->spinlock, flags);
>> +    /*
>> +     * Since the sysfs buffer allocation and the hardware enablement 
>> is not
>> +     * in the same critical region, it's possible to race with the perf
>> +     * mode:
>> +     *   [sysfs mode]                   [perf mode]
>> +     *   tmc_etr_get_sysfs_buffer()
>> +     *     spin_lock(&drvdata->spinlock)
>> +     *     [sysfs buffer allocation]
>> +     *     spin_unlock(&drvdata->spinlock)
>> +     *                                  spin_lock(&drvdata->spinlock)
>> +     *                                  tmc_etr_enable_hw()
>> +     *                                    drvdata->etr_buf = 
>> etr_perf->etr_buf
>> +     *                                  spin_unlock(&drvdata->spinlock)
>> +     *   spin_lock(&drvdata->spinlock)
>> +     *   tmc_etr_enable_hw()
>> +     *     WARN_ON(drvdata->etr_buf) // WARN sicne etr_buf 
>> initialized at
>> +     *                                  the perf side
>> +     *   spin_unlock(&drvdata->spinlock)
>> +     *
>> +     * So check here before continue.
>> +     */
>> +    if (coresight_get_mode(csdev) == CS_MODE_PERF) {
>> +        drvdata->sysfs_buf = NULL;
>> +        spin_unlock_irqrestore(&drvdata->spinlock, flags);
>> +
>> +        /* Free allocated memory out side of the spinlock */
>> +        tmc_etr_free_sysfs_buf(sysfs_buf);
>> +        return -EBUSY;
>> +    }
> 
> With this in place, I think we should remove the check for CS_MODE_PERF 
> in get_etr_sysfs_buf() to avoid confusion (which I believe opened up 
> this race)
> 
> Suzuki
> 

Is it not simpler to set the mode to SYSFS before allocating the buffer 
in the first place? Then we don't need to free if it races and can't get 
into the intermediate state where it's a half initialized sysfs mode. 
The lock doesn't need to be held the whole time, just when setting the mode.

Or maybe to make it more consistent with etm4_enable() use 
coresight_take_mode() outside of the lock.

And then also clean up the perf mode check in get_etr_sysfs_buf().

diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c 
b/drivers/hwtracing/coresight/coresight-tmc-etr.c
index a48bb85d0e7f..29c07832127b 100644
--- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
+++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
@@ -1219,13 +1219,17 @@ static int tmc_enable_etr_sink_sysfs(struct 
coresight_device *csdev)
         int ret = 0;
         unsigned long flags;
         struct tmc_drvdata *drvdata = dev_get_drvdata(csdev->dev.parent);
-       struct etr_buf *sysfs_buf = tmc_etr_get_sysfs_buffer(csdev);
+       struct etr_buf *sysfs_buf;

+       spin_lock_irqsave(&drvdata->spinlock, flags);
+
+       if (coresight_get_mode(csdev) == CS_MODE_PERF)
+               return -EBUSY;
+
+       sysfs_buf = tmc_etr_get_sysfs_buffer(csdev);
         if (IS_ERR(sysfs_buf))
                 return PTR_ERR(sysfs_buf);

-       spin_lock_irqsave(&drvdata->spinlock, flags);
-
         /*
          * In sysFS mode we can have multiple writers per sink.  Since this
          * sink is already enabled no memory is needed and the HW need 
not be
Suzuki K Poulose Dec. 2, 2024, 11:31 a.m. UTC | #4
On 02/12/2024 11:19, James Clark wrote:
> 
> 
> On 02/12/2024 9:46 am, Suzuki K Poulose wrote:
>> Hi Yicong
>>
>> On 02/12/2024 09:24, Yicong Yang wrote:
>>> From: Yicong Yang <yangyicong@hisilicon.com>
>>>
>>> When trying to run perf and sysfs mode simultaneously, the WARN_ON()
>>> in tmc_etr_enable_hw() is triggered sometimes:
>>>
>>>   WARNING: CPU: 42 PID: 3911571 at drivers/hwtracing/coresight/ 
>>> coresight-tmc-etr.c:1060 tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc]
>>>   [..snip..]
>>>   Call trace:
>>>    tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc] (P)
>>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc] (L)
>>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc]
>>>    coresight_enable_path+0x1c8/0x218 [coresight]
>>>    coresight_enable_sysfs+0xa4/0x228 [coresight]
>>>    enable_source_store+0x58/0xa8 [coresight]
>>>    dev_attr_store+0x20/0x40
>>>    sysfs_kf_write+0x4c/0x68
>>>    kernfs_fop_write_iter+0x120/0x1b8
>>>    vfs_write+0x2c8/0x388
>>>    ksys_write+0x74/0x108
>>>    __arm64_sys_write+0x24/0x38
>>>    el0_svc_common.constprop.0+0x64/0x148
>>>    do_el0_svc+0x24/0x38
>>>    el0_svc+0x3c/0x130
>>>    el0t_64_sync_handler+0xc8/0xd0
>>>    el0t_64_sync+0x1ac/0x1b0
>>>   ---[ end trace 0000000000000000 ]---
>>>
>>> Since the enablement of sysfs mode is separeted into two critical 
>>> regions,
>>> one for sysfs buffer allocation and another for hardware enablement, 
>>> it's
>>> possible to race with the perf mode. Fix this by double check whether
>>> the perf mode's been used before enabling the hardware in sysfs mode.
>>
>> Thanks for the fix. Some minor comments below.
>>
>> It needs a Fixes tag.
>>
>>>
>>> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
>>> ---
>>>   .../hwtracing/coresight/coresight-tmc-etr.c   | 30 +++++++++++++++++++
>>>   1 file changed, 30 insertions(+)
>>>
>>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/ 
>>> drivers/hwtracing/coresight/coresight-tmc-etr.c
>>> index ad83714ca4dc..d382d95da5ff 100644
>>> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
>>> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
>>> @@ -1230,6 +1230,36 @@ static int tmc_enable_etr_sink_sysfs(struct 
>>> coresight_device *csdev)
>>>       spin_lock_irqsave(&drvdata->spinlock, flags);
>>> +    /*
>>> +     * Since the sysfs buffer allocation and the hardware enablement 
>>> is not
>>> +     * in the same critical region, it's possible to race with the perf
>>> +     * mode:
>>> +     *   [sysfs mode]                   [perf mode]
>>> +     *   tmc_etr_get_sysfs_buffer()
>>> +     *     spin_lock(&drvdata->spinlock)
>>> +     *     [sysfs buffer allocation]
>>> +     *     spin_unlock(&drvdata->spinlock)
>>> +     *                                  spin_lock(&drvdata->spinlock)
>>> +     *                                  tmc_etr_enable_hw()
>>> +     *                                    drvdata->etr_buf = 
>>> etr_perf->etr_buf
>>> +     *                                  spin_unlock(&drvdata->spinlock)
>>> +     *   spin_lock(&drvdata->spinlock)
>>> +     *   tmc_etr_enable_hw()
>>> +     *     WARN_ON(drvdata->etr_buf) // WARN sicne etr_buf 
>>> initialized at
>>> +     *                                  the perf side
>>> +     *   spin_unlock(&drvdata->spinlock)
>>> +     *
>>> +     * So check here before continue.
>>> +     */
>>> +    if (coresight_get_mode(csdev) == CS_MODE_PERF) {
>>> +        drvdata->sysfs_buf = NULL;
>>> +        spin_unlock_irqrestore(&drvdata->spinlock, flags);
>>> +
>>> +        /* Free allocated memory out side of the spinlock */
>>> +        tmc_etr_free_sysfs_buf(sysfs_buf);
>>> +        return -EBUSY;
>>> +    }
>>
>> With this in place, I think we should remove the check for 
>> CS_MODE_PERF in get_etr_sysfs_buf() to avoid confusion (which I 
>> believe opened up this race)
>>
>> Suzuki
>>
> 
> Is it not simpler to set the mode to SYSFS before allocating the buffer 
> in the first place? Then we don't need to free if it races and can't get 
> into the intermediate state where it's a half initialized sysfs mode. 
> The lock doesn't need to be held the whole time, just when setting the 
> mode.
> 
> Or maybe to make it more consistent with etm4_enable() use 
> coresight_take_mode() outside of the lock.
> 
> And then also clean up the perf mode check in get_etr_sysfs_buf().
> 
> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/drivers/ 
> hwtracing/coresight/coresight-tmc-etr.c
> index a48bb85d0e7f..29c07832127b 100644
> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
> @@ -1219,13 +1219,17 @@ static int tmc_enable_etr_sink_sysfs(struct 
> coresight_device *csdev)
>          int ret = 0;
>          unsigned long flags;
>          struct tmc_drvdata *drvdata = dev_get_drvdata(csdev->dev.parent);
> -       struct etr_buf *sysfs_buf = tmc_etr_get_sysfs_buffer(csdev);
> +       struct etr_buf *sysfs_buf;
> 
> +       spin_lock_irqsave(&drvdata->spinlock, flags);
> +
> +       if (coresight_get_mode(csdev) == CS_MODE_PERF)
> +               return -EBUSY;
> +
> +       sysfs_buf = tmc_etr_get_sysfs_buffer(csdev);

^^ This would try to loack the spinlock again. Or we should explicitly 
assert that the spinlock is held in the tmc_etr_get_sysfs_buffer() and 
release it and then allocate the buffer. Which again opens up a race
for a PERF session to take it over ?

Suzuki


>          if (IS_ERR(sysfs_buf))
>                  return PTR_ERR(sysfs_buf);
> 
> -       spin_lock_irqsave(&drvdata->spinlock, flags);
> -
>          /*
>           * In sysFS mode we can have multiple writers per sink.  Since 
> this
>           * sink is already enabled no memory is needed and the HW need 
> not be
> 
>
Suzuki K Poulose Dec. 2, 2024, 11:38 a.m. UTC | #5
On 02/12/2024 11:31, Suzuki K Poulose wrote:
> On 02/12/2024 11:19, James Clark wrote:
>>
>>
>> On 02/12/2024 9:46 am, Suzuki K Poulose wrote:
>>> Hi Yicong
>>>
>>> On 02/12/2024 09:24, Yicong Yang wrote:
>>>> From: Yicong Yang <yangyicong@hisilicon.com>
>>>>
>>>> When trying to run perf and sysfs mode simultaneously, the WARN_ON()
>>>> in tmc_etr_enable_hw() is triggered sometimes:
>>>>
>>>>   WARNING: CPU: 42 PID: 3911571 at drivers/hwtracing/coresight/ 
>>>> coresight-tmc-etr.c:1060 tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc]
>>>>   [..snip..]
>>>>   Call trace:
>>>>    tmc_etr_enable_hw+0xc0/0xd8 [coresight_tmc] (P)
>>>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc] (L)
>>>>    tmc_enable_etr_sink+0x11c/0x250 [coresight_tmc]
>>>>    coresight_enable_path+0x1c8/0x218 [coresight]
>>>>    coresight_enable_sysfs+0xa4/0x228 [coresight]
>>>>    enable_source_store+0x58/0xa8 [coresight]
>>>>    dev_attr_store+0x20/0x40
>>>>    sysfs_kf_write+0x4c/0x68
>>>>    kernfs_fop_write_iter+0x120/0x1b8
>>>>    vfs_write+0x2c8/0x388
>>>>    ksys_write+0x74/0x108
>>>>    __arm64_sys_write+0x24/0x38
>>>>    el0_svc_common.constprop.0+0x64/0x148
>>>>    do_el0_svc+0x24/0x38
>>>>    el0_svc+0x3c/0x130
>>>>    el0t_64_sync_handler+0xc8/0xd0
>>>>    el0t_64_sync+0x1ac/0x1b0
>>>>   ---[ end trace 0000000000000000 ]---
>>>>
>>>> Since the enablement of sysfs mode is separeted into two critical 
>>>> regions,
>>>> one for sysfs buffer allocation and another for hardware enablement, 
>>>> it's
>>>> possible to race with the perf mode. Fix this by double check whether
>>>> the perf mode's been used before enabling the hardware in sysfs mode.
>>>
>>> Thanks for the fix. Some minor comments below.
>>>
>>> It needs a Fixes tag.
>>>
>>>>
>>>> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
>>>> ---
>>>>   .../hwtracing/coresight/coresight-tmc-etr.c   | 30 +++++++++++++++ 
>>>> ++++
>>>>   1 file changed, 30 insertions(+)
>>>>
>>>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/ 
>>>> drivers/hwtracing/coresight/coresight-tmc-etr.c
>>>> index ad83714ca4dc..d382d95da5ff 100644
>>>> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
>>>> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
>>>> @@ -1230,6 +1230,36 @@ static int tmc_enable_etr_sink_sysfs(struct 
>>>> coresight_device *csdev)
>>>>       spin_lock_irqsave(&drvdata->spinlock, flags);
>>>> +    /*
>>>> +     * Since the sysfs buffer allocation and the hardware 
>>>> enablement is not
>>>> +     * in the same critical region, it's possible to race with the 
>>>> perf
>>>> +     * mode:
>>>> +     *   [sysfs mode]                   [perf mode]
>>>> +     *   tmc_etr_get_sysfs_buffer()
>>>> +     *     spin_lock(&drvdata->spinlock)
>>>> +     *     [sysfs buffer allocation]
>>>> +     *     spin_unlock(&drvdata->spinlock)
>>>> +     *                                  spin_lock(&drvdata->spinlock)
>>>> +     *                                  tmc_etr_enable_hw()
>>>> +     *                                    drvdata->etr_buf = 
>>>> etr_perf->etr_buf
>>>> +     *                                  spin_unlock(&drvdata- 
>>>> >spinlock)
>>>> +     *   spin_lock(&drvdata->spinlock)
>>>> +     *   tmc_etr_enable_hw()
>>>> +     *     WARN_ON(drvdata->etr_buf) // WARN sicne etr_buf 
>>>> initialized at
>>>> +     *                                  the perf side
>>>> +     *   spin_unlock(&drvdata->spinlock)
>>>> +     *
>>>> +     * So check here before continue.
>>>> +     */
>>>> +    if (coresight_get_mode(csdev) == CS_MODE_PERF) {
>>>> +        drvdata->sysfs_buf = NULL;
>>>> +        spin_unlock_irqrestore(&drvdata->spinlock, flags);
>>>> +
>>>> +        /* Free allocated memory out side of the spinlock */
>>>> +        tmc_etr_free_sysfs_buf(sysfs_buf);
>>>> +        return -EBUSY;
>>>> +    }
>>>
>>> With this in place, I think we should remove the check for 
>>> CS_MODE_PERF in get_etr_sysfs_buf() to avoid confusion (which I 
>>> believe opened up this race)
>>>
>>> Suzuki
>>>
>>
>> Is it not simpler to set the mode to SYSFS before allocating the 
>> buffer in the first place? Then we don't need to free if it races and 
>> can't get into the intermediate state where it's a half initialized 
>> sysfs mode. The lock doesn't need to be held the whole time, just when 
>> setting the mode.
>>
>> Or maybe to make it more consistent with etm4_enable() use 
>> coresight_take_mode() outside of the lock.
>>
>> And then also clean up the perf mode check in get_etr_sysfs_buf().
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/ 
>> drivers/ hwtracing/coresight/coresight-tmc-etr.c
>> index a48bb85d0e7f..29c07832127b 100644
>> --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
>> @@ -1219,13 +1219,17 @@ static int tmc_enable_etr_sink_sysfs(struct 
>> coresight_device *csdev)
>>          int ret = 0;
>>          unsigned long flags;
>>          struct tmc_drvdata *drvdata = dev_get_drvdata(csdev- 
>> >dev.parent);
>> -       struct etr_buf *sysfs_buf = tmc_etr_get_sysfs_buffer(csdev);
>> +       struct etr_buf *sysfs_buf;
>>
>> +       spin_lock_irqsave(&drvdata->spinlock, flags);
>> +
>> +       if (coresight_get_mode(csdev) == CS_MODE_PERF)
>> +               return -EBUSY;
>> +
>> +       sysfs_buf = tmc_etr_get_sysfs_buffer(csdev);
> 
> ^^ This would try to loack the spinlock again. Or we should explicitly 
> assert that the spinlock is held in the tmc_etr_get_sysfs_buffer() and 
> release it and then allocate the buffer. Which again opens up a race
> for a PERF session to take it over ?

Ah, no, the mode will prevent PERF mode taking over.

Suzuki

> 
> Suzuki
> 
> 
>>          if (IS_ERR(sysfs_buf))
>>                  return PTR_ERR(sysfs_buf);
>>
>> -       spin_lock_irqsave(&drvdata->spinlock, flags);
>> -
>>          /*
>>           * In sysFS mode we can have multiple writers per sink.  
>> Since this
>>           * sink is already enabled no memory is needed and the HW 
>> need not be
>>
>>
>
diff mbox series

Patch

diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/drivers/hwtracing/coresight/coresight-tmc-etr.c
index ad83714ca4dc..d382d95da5ff 100644
--- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
+++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
@@ -1230,6 +1230,36 @@  static int tmc_enable_etr_sink_sysfs(struct coresight_device *csdev)
 
 	spin_lock_irqsave(&drvdata->spinlock, flags);
 
+	/*
+	 * Since the sysfs buffer allocation and the hardware enablement is not
+	 * in the same critical region, it's possible to race with the perf
+	 * mode:
+	 *   [sysfs mode]                   [perf mode]
+	 *   tmc_etr_get_sysfs_buffer()
+	 *     spin_lock(&drvdata->spinlock)
+	 *     [sysfs buffer allocation]
+	 *     spin_unlock(&drvdata->spinlock)
+	 *                                  spin_lock(&drvdata->spinlock)
+	 *                                  tmc_etr_enable_hw()
+	 *                                    drvdata->etr_buf = etr_perf->etr_buf
+	 *                                  spin_unlock(&drvdata->spinlock)
+	 *   spin_lock(&drvdata->spinlock)
+	 *   tmc_etr_enable_hw()
+	 *     WARN_ON(drvdata->etr_buf) // WARN sicne etr_buf initialized at
+	 *                                  the perf side
+	 *   spin_unlock(&drvdata->spinlock)
+	 *
+	 * So check here before continue.
+	 */
+	if (coresight_get_mode(csdev) == CS_MODE_PERF) {
+		drvdata->sysfs_buf = NULL;
+		spin_unlock_irqrestore(&drvdata->spinlock, flags);
+
+		/* Free allocated memory out side of the spinlock */
+		tmc_etr_free_sysfs_buf(sysfs_buf);
+		return -EBUSY;
+	}
+
 	/*
 	 * In sysFS mode we can have multiple writers per sink.  Since this
 	 * sink is already enabled no memory is needed and the HW need not be