[GIT PULL v1 0/7] stm class/intel_th: Updates for v5.1

[GIT PULL v1 0/7] stm class/intel_th: Updates for v5.1

[Alexander Shishkin]

Hi Greg,

Second attempt. Added stable CC and Fixes tags to fixes.

Here are the STM and Intel TH changes that I have for v5.1. Nothing big,
a few fixes that are too late for the v5.0 and a few minor cleanups and
a documentation update. All sparse/smatch/coccinelle/cppcheck/checkpatch
clean as usual, courtesy of aiaiai.
Individual patches follow. Please consider pulling or applying. Thanks!

The following changes since commit bfeffd155283772bbe78c6a05dec7c0128ee500c:

  Linux 5.0-rc1 (2019-01-06 17:08:20 -0800)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/ash/stm.git tags/intel_th-stm-for-greg-20190221

for you to fetch changes up to 4c3cef46a96e8f623d1edf03424880ebc901b068:

  stm class: Prevent division by zero (2019-02-21 15:29:26 +0200)

----------------------------------------------------------------
stm class/intel_th: Updates for v5.1

These are:
  * 2 bugfixes in stm class
  * one bugfix in intel_th
  * a few minor cleanups

----------------------------------------------------------------
Alexander Shishkin (4):
  intel_th: Update ABI documentation
  intel_th: Only create useful device nodes
  intel_th: Don't reference unassigned outputs
  stm class: Prevent division by zero

Andy Shevchenko (1):
  intel_th: pti: Use sysfs_match_string() helper

Gustavo A. R. Silva (1):
  intel_th: Mark expected switch fall-throughs

Zhi Jin (1):
  stm class: Fix an endless loop in channel allocation

 .../testing/sysfs-bus-intel_th-output-devices    |  6 ++++--
 drivers/hwtracing/intel_th/core.c                |  6 +++++-
 drivers/hwtracing/intel_th/gth.c                 |  4 ++++
 drivers/hwtracing/intel_th/pti.c                 | 16 +++++++---------
 drivers/hwtracing/intel_th/sth.c                 |  4 ++++
 drivers/hwtracing/stm/core.c                     | 11 ++++++++---
 6 files changed, 32 insertions(+), 15 deletions(-)

-- 
2.20.1

[GIT PULL v1 1/7] intel_th: Update ABI documentation

[Alexander Shishkin]

Commit a753bfcfdb1f3 ("intel_th: Make the switch allocate its subdevices")
changed the behavior so that the output port devices are created only for
the ports reported by the hardware and their initial state is "unassigned"
until a corresponding output port driver is loaded. Reflect this fact in
the ABI documentation.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reported-by: Ricardo Neri <ricardo.neri@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-intel_th-output-devices | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-bus-intel_th-output-devices b/Documentation/ABI/testing/sysfs-bus-intel_th-output-devices
index 4d48a9451866..d1f667104944 100644
--- a/Documentation/ABI/testing/sysfs-bus-intel_th-output-devices
+++ b/Documentation/ABI/testing/sysfs-bus-intel_th-output-devices
@@ -3,11 +3,13 @@ Date:		June 2015
 KernelVersion:	4.3
 Contact:	Alexander Shishkin <alexander.shishkin@linux.intel.com>
 Description:	(RW) Writes of 1 or 0 enable or disable trace output to this
-		output device. Reads return current status.
+		output device. Reads return current status. Requires that the
+		correstponding output port driver be loaded.
 
 What:		/sys/bus/intel_th/devices/<intel_th_id>-msc<msc-id>/port
 Date:		June 2015
 KernelVersion:	4.3
 Contact:	Alexander Shishkin <alexander.shishkin@linux.intel.com>
 Description:	(RO) Port number, corresponding to this output device on the
-		switch (GTH).
+		switch (GTH) or "unassigned" if the corresponding output
+		port driver is not loaded.
-- 
2.20.1

[GIT PULL v1 2/7] intel_th: Mark expected switch fall-throughs

[Alexander Shishkin]

From: "Gustavo A. R. Silva" <gustavo@embeddedor.com>

In preparation to enabling -Wimplicit-fallthrough, mark switch
cases where we are expecting to fall through.

This patch fixes the following warnings:

drivers/hwtracing/intel_th/sth.c: In function ‘sth_stm_packet’:
drivers/hwtracing/intel_th/sth.c:86:7: warning: this statement may fall through [-Wimplicit-fallthrough=]
   reg += 4;
   ~~~~^~~~
drivers/hwtracing/intel_th/sth.c:87:2: note: here
  case STP_PACKET_XSYNC:
  ^~~~
drivers/hwtracing/intel_th/sth.c:88:7: warning: this statement may fall through [-Wimplicit-fallthrough=]
   reg += 8;
   ~~~~^~~~
drivers/hwtracing/intel_th/sth.c:89:2: note: here
  case STP_PACKET_TRIG:
  ^~~~

Warning level 3 was used: -Wimplicit-fallthrough=3

This patch is part of the ongoing efforts to enable
-Wimplicit-fallthrough.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
---
 drivers/hwtracing/intel_th/sth.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/hwtracing/intel_th/sth.c b/drivers/hwtracing/intel_th/sth.c
index 4b7ae47789d2..3a1f4e650378 100644
--- a/drivers/hwtracing/intel_th/sth.c
+++ b/drivers/hwtracing/intel_th/sth.c
@@ -84,8 +84,12 @@ static ssize_t notrace sth_stm_packet(struct stm_data *stm_data,
 	/* Global packets (GERR, XSYNC, TRIG) are sent with register writes */
 	case STP_PACKET_GERR:
 		reg += 4;
+		/* fall through */
+
 	case STP_PACKET_XSYNC:
 		reg += 8;
+		/* fall through */
+
 	case STP_PACKET_TRIG:
 		if (flags & STP_PACKET_TIMESTAMPED)
 			reg += 4;
-- 
2.20.1

[GIT PULL v1 3/7] intel_th: Only create useful device nodes

[Alexander Shishkin]

Right now, the driver will create a device node for each output port,
with the intent to provide read access to that port's data. However,
only the memory ports are readable this way (msc0, msc1). Other output
ports don't need device nodes, so remove them.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
---
 drivers/hwtracing/intel_th/core.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/hwtracing/intel_th/core.c b/drivers/hwtracing/intel_th/core.c
index fc6b7f8b62fb..7c1acc2f801c 100644
--- a/drivers/hwtracing/intel_th/core.c
+++ b/drivers/hwtracing/intel_th/core.c
@@ -422,6 +422,7 @@ static const struct intel_th_subdevice {
 	unsigned		nres;
 	unsigned		type;
 	unsigned		otype;
+	bool			mknode;
 	unsigned		scrpd;
 	int			id;
 } intel_th_subdevices[] = {
@@ -456,6 +457,7 @@ static const struct intel_th_subdevice {
 		.name	= "msc",
 		.id	= 0,
 		.type	= INTEL_TH_OUTPUT,
+		.mknode	= true,
 		.otype	= GTH_MSU,
 		.scrpd	= SCRPD_MEM_IS_PRIM_DEST | SCRPD_MSC0_IS_ENABLED,
 	},
@@ -476,6 +478,7 @@ static const struct intel_th_subdevice {
 		.name	= "msc",
 		.id	= 1,
 		.type	= INTEL_TH_OUTPUT,
+		.mknode	= true,
 		.otype	= GTH_MSU,
 		.scrpd	= SCRPD_MEM_IS_PRIM_DEST | SCRPD_MSC1_IS_ENABLED,
 	},
@@ -635,7 +638,8 @@ intel_th_subdevice_alloc(struct intel_th *th,
 	}
 
 	if (subdev->type == INTEL_TH_OUTPUT) {
-		thdev->dev.devt = MKDEV(th->major, th->num_thdevs);
+		if (subdev->mknode)
+			thdev->dev.devt = MKDEV(th->major, th->num_thdevs);
 		thdev->output.type = subdev->otype;
 		thdev->output.port = -1;
 		thdev->output.scratchpad = subdev->scrpd;
-- 
2.20.1

[GIT PULL v1 4/7] intel_th: pti: Use sysfs_match_string() helper

[Alexander Shishkin]

From: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

Use sysfs_match_string() helper instead of open coded variant.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
---
 drivers/hwtracing/intel_th/pti.c | 16 +++++++---------
 1 file changed, 7 insertions(+), 9 deletions(-)

diff --git a/drivers/hwtracing/intel_th/pti.c b/drivers/hwtracing/intel_th/pti.c
index 56694339cb06..0da6b787f553 100644
--- a/drivers/hwtracing/intel_th/pti.c
+++ b/drivers/hwtracing/intel_th/pti.c
@@ -272,19 +272,17 @@ static ssize_t lpp_dest_store(struct device *dev, struct device_attribute *attr,
 			      const char *buf, size_t size)
 {
 	struct pti_device *pti = dev_get_drvdata(dev);
-	ssize_t ret = -EINVAL;
 	int i;
 
-	for (i = 0; i < ARRAY_SIZE(lpp_dest_str); i++)
-		if (sysfs_streq(buf, lpp_dest_str[i]))
-			break;
+	i = sysfs_match_string(lpp_dest_str, buf);
+	if (i < 0)
+		return i;
 
-	if (i < ARRAY_SIZE(lpp_dest_str) && pti->lpp_dest_mask & BIT(i)) {
-		pti->lpp_dest = i;
-		ret = size;
-	}
+	if (!(pti->lpp_dest_mask & BIT(i)))
+		return -EINVAL;
 
-	return ret;
+	pti->lpp_dest = i;
+	return size;
 }
 
 static DEVICE_ATTR_RW(lpp_dest);
-- 
2.20.1

[GIT PULL v1 5/7] intel_th: Don't reference unassigned outputs

[Alexander Shishkin]

When an output port driver is removed, also remove references to it from
any masters. Failing to do this causes a NULL ptr dereference when
configuring another output port:

> BUG: unable to handle kernel NULL pointer dereference at 000000000000000d
> RIP: 0010:master_attr_store+0x9d/0x160 [intel_th_gth]
> Call Trace:
> dev_attr_store+0x1b/0x30
> sysfs_kf_write+0x3c/0x50
> kernfs_fop_write+0x125/0x1a0
> __vfs_write+0x3a/0x190
> ? __vfs_write+0x5/0x190
> ? _cond_resched+0x1a/0x50
> ? rcu_all_qs+0x5/0xb0
> ? __vfs_write+0x5/0x190
> vfs_write+0xb8/0x1b0
> ksys_write+0x55/0xc0
> __x64_sys_write+0x1a/0x20
> do_syscall_64+0x5a/0x140
> entry_SYSCALL_64_after_hwframe+0x44/0xa9

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Fixes: b27a6a3f97b9 ("intel_th: Add Global Trace Hub driver")
CC: stable@vger.kernel.org # v4.4+
Reported-by: Ammy Yi <ammy.yi@intel.com>
---
 drivers/hwtracing/intel_th/gth.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/hwtracing/intel_th/gth.c b/drivers/hwtracing/intel_th/gth.c
index 8426b7970c14..cc287cf6eb29 100644
--- a/drivers/hwtracing/intel_th/gth.c
+++ b/drivers/hwtracing/intel_th/gth.c
@@ -607,6 +607,7 @@ static void intel_th_gth_unassign(struct intel_th_device *thdev,
 {
 	struct gth_device *gth = dev_get_drvdata(&thdev->dev);
 	int port = othdev->output.port;
+	int master;
 
 	if (thdev->host_mode)
 		return;
@@ -615,6 +616,9 @@ static void intel_th_gth_unassign(struct intel_th_device *thdev,
 	othdev->output.port = -1;
 	othdev->output.active = false;
 	gth->output[port].output = NULL;
+	for (master = 0; master < TH_CONFIGURABLE_MASTERS; master++)
+		if (gth->master[master] == port)
+			gth->master[master] = -1;
 	spin_unlock(&gth->gth_lock);
 }
 
-- 
2.20.1

[GIT PULL v1 6/7] stm class: Fix an endless loop in channel allocation

[Alexander Shishkin]

From: Zhi Jin <zhi.jin@intel.com>

There is a bug in the channel allocation logic that leads to an endless
loop when looking for a contiguous range of channels in a range with a
mixture of free and occupied channels. For example, opening three
consequtive channels, closing the first two and requesting 4 channels in
a row will trigger this soft lockup. The bug is that the search loop
forgets to skip over the range once it detects that one channel in that
range is occupied.

Restore the original intent to the logic by fixing the omission.

Signed-off-by: Zhi Jin <zhi.jin@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Fixes: 7bd1d4093c2f ("stm class: Introduce an abstraction for System Trace Module devices")
CC: stable@vger.kernel.org # v4.4+
---
 drivers/hwtracing/stm/core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/hwtracing/stm/core.c b/drivers/hwtracing/stm/core.c
index 93ce3aa740a9..c80b064224f6 100644
--- a/drivers/hwtracing/stm/core.c
+++ b/drivers/hwtracing/stm/core.c
@@ -244,6 +244,9 @@ static int find_free_channels(unsigned long *bitmap, unsigned int start,
 			;
 		if (i == width)
 			return pos;
+
+		/* step over [pos..pos+i) to continue search */
+		pos += i;
 	}
 
 	return -1;
-- 
2.20.1

[GIT PULL v1 7/7] stm class: Prevent division by zero

[Alexander Shishkin]

Using STP_POLICY_ID_SET ioctl command with dummy_stm device, or any STM
device that supplies zero mmio channel size, will trigger a division by
zero bug in the kernel.

Prevent this by disallowing channel widths other than 1 for such devices.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Fixes: 7bd1d4093c2f ("stm class: Introduce an abstraction for System Trace Module devices")
CC: stable@vger.kernel.org # v4.4+
---
 drivers/hwtracing/stm/core.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/hwtracing/stm/core.c b/drivers/hwtracing/stm/core.c
index c80b064224f6..c7ba8acfd4d5 100644
--- a/drivers/hwtracing/stm/core.c
+++ b/drivers/hwtracing/stm/core.c
@@ -735,7 +735,7 @@ static int stm_char_policy_set_ioctl(struct stm_file *stmf, void __user *arg)
 	struct stm_device *stm = stmf->stm;
 	struct stp_policy_id *id;
 	char *ids[] = { NULL, NULL };
-	int ret = -EINVAL;
+	int ret = -EINVAL, wlimit = 1;
 	u32 size;
 
 	if (stmf->output.nr_chans)
@@ -763,8 +763,10 @@ static int stm_char_policy_set_ioctl(struct stm_file *stmf, void __user *arg)
 	if (id->__reserved_0 || id->__reserved_1)
 		goto err_free;
 
-	if (id->width < 1 ||
-	    id->width > PAGE_SIZE / stm->data->sw_mmiosz)
+	if (stm->data->sw_mmiosz)
+		wlimit = PAGE_SIZE / stm->data->sw_mmiosz;
+
+	if (id->width < 1 || id->width > wlimit)
 		goto err_free;
 
 	ids[0] = id->id;
-- 
2.20.1

Re: [GIT PULL v1 0/7] stm class/intel_th: Updates for v5.1

[Greg Kroah-Hartman]

On Thu, Feb 21, 2019 at 05:49:59PM +0200, Alexander Shishkin wrote:
> Hi Greg,
> 
> Second attempt. Added stable CC and Fixes tags to fixes.
> 
> Here are the STM and Intel TH changes that I have for v5.1. Nothing big,
> a few fixes that are too late for the v5.0 and a few minor cleanups and
> a documentation update. All sparse/smatch/coccinelle/cppcheck/checkpatch
> clean as usual, courtesy of aiaiai.
> Individual patches follow. Please consider pulling or applying. Thanks!
> 
> The following changes since commit bfeffd155283772bbe78c6a05dec7c0128ee500c:
> 
>   Linux 5.0-rc1 (2019-01-06 17:08:20 -0800)
> 
> are available in the Git repository at:
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/ash/stm.git tags/intel_th-stm-for-greg-20190221

Much better, thanks, pulled and pushed out.

greg k-h

Re: [GIT PULL v1 0/7] stm class/intel_th: Updates for v5.1

[Alexander Shishkin]

Greg Kroah-Hartman <gregkh@linuxfoundation.org> writes:

> On Thu, Feb 21, 2019 at 05:49:59PM +0200, Alexander Shishkin wrote:
>> Hi Greg,
>> 
>> Second attempt. Added stable CC and Fixes tags to fixes.
>> 
>> Here are the STM and Intel TH changes that I have for v5.1. Nothing big,
>> a few fixes that are too late for the v5.0 and a few minor cleanups and
>> a documentation update. All sparse/smatch/coccinelle/cppcheck/checkpatch
>> clean as usual, courtesy of aiaiai.
>> Individual patches follow. Please consider pulling or applying. Thanks!
>> 
>> The following changes since commit bfeffd155283772bbe78c6a05dec7c0128ee500c:
>> 
>>   Linux 5.0-rc1 (2019-01-06 17:08:20 -0800)
>> 
>> are available in the Git repository at:
>> 
>>   git://git.kernel.org/pub/scm/linux/kernel/git/ash/stm.git tags/intel_th-stm-for-greg-20190221
>
> Much better, thanks, pulled and pushed out.

This is super embarrassing, but 5/7 in this set has an off-by-one. A fix
is headed to your inbox [1]. Will do better next time!

[1] https://marc.info/?l=linux-kernel&m=155142781026182

Thanks,
--
Alex