Before 98f4c651762c, we returned zeros for unopened channels.
With 98f4c651762c, we started to return random on-stack values.
We'd better return -EINVAL instead.
Fixes: 98f4c651762c ("hv: move ringbuffer bus attributes to dev_groups")
Cc: [email protected]
Cc: K. Y. Srinivasan <[email protected]>
Cc: Haiyang Zhang <[email protected]>
Cc: Stephen Hemminger <[email protected]>
Signed-off-by: Dexuan Cui <[email protected]>
---
drivers/hv/vmbus_drv.c | 20 ++++++++++++++++++++
1 file changed, 20 insertions(+)
diff --git a/drivers/hv/vmbus_drv.c b/drivers/hv/vmbus_drv.c
index 283d184..d0ff656 100644
--- a/drivers/hv/vmbus_drv.c
+++ b/drivers/hv/vmbus_drv.c
@@ -316,6 +316,8 @@ static ssize_t out_intr_mask_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
return sprintf(buf, "%d\n", outbound.current_interrupt_mask);
}
@@ -329,6 +331,8 @@ static ssize_t out_read_index_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
return sprintf(buf, "%d\n", outbound.current_read_index);
}
@@ -343,6 +347,8 @@ static ssize_t out_write_index_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
return sprintf(buf, "%d\n", outbound.current_write_index);
}
@@ -357,6 +363,8 @@ static ssize_t out_read_bytes_avail_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
return sprintf(buf, "%d\n", outbound.bytes_avail_toread);
}
@@ -371,6 +379,8 @@ static ssize_t out_write_bytes_avail_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
return sprintf(buf, "%d\n", outbound.bytes_avail_towrite);
}
@@ -384,6 +394,8 @@ static ssize_t in_intr_mask_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
return sprintf(buf, "%d\n", inbound.current_interrupt_mask);
}
@@ -397,6 +409,8 @@ static ssize_t in_read_index_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
return sprintf(buf, "%d\n", inbound.current_read_index);
}
@@ -410,6 +424,8 @@ static ssize_t in_write_index_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
return sprintf(buf, "%d\n", inbound.current_write_index);
}
@@ -424,6 +440,8 @@ static ssize_t in_read_bytes_avail_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
return sprintf(buf, "%d\n", inbound.bytes_avail_toread);
}
@@ -438,6 +456,8 @@ static ssize_t in_write_bytes_avail_show(struct device *dev,
if (!hv_dev->channel)
return -ENODEV;
+ if (hv_dev->channel->state != CHANNEL_OPENED_STATE)
+ return -EINVAL;
hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
return sprintf(buf, "%d\n", inbound.bytes_avail_towrite);
}
--
2.7.4
On Thu, Dec 13, 2018 at 04:35:43PM +0000, Dexuan Cui wrote:
>
>Before 98f4c651762c, we returned zeros for unopened channels.
>With 98f4c651762c, we started to return random on-stack values.
>
>We'd better return -EINVAL instead.
>
>Fixes: 98f4c651762c ("hv: move ringbuffer bus attributes to dev_groups")
>Cc: [email protected]
>Cc: K. Y. Srinivasan <[email protected]>
>Cc: Haiyang Zhang <[email protected]>
>Cc: Stephen Hemminger <[email protected]>
>Signed-off-by: Dexuan Cui <[email protected]>
Queued up, thank you.
--
Thanks,
Sasha
> From: Stephen Hemminger <[email protected]>
> On Thu, 13 Dec 2018 16:35:43 +0000
> Dexuan Cui <[email protected]> wrote:
>
> > Before 98f4c651762c, we returned zeros for unopened channels.
> > With 98f4c651762c, we started to return random on-stack values.
> >
> > We'd better return -EINVAL instead.
>
> The concept looks fine, but maybe it would be simpler to move it into
> hv_ringbuffer_get_debuginfo and have it return an error code.
>
> Since so much of the code is repeated, I would probably make a
> macro which generates the code as well.
>
> Something like this:
Thanks, Stephen! Now the patch has been in char-misc's char-misc-linus
branch, so IMO we may as well leave it as is (considering the code here is
unlikely to be frqeuencly changed), and we have a smaller patch this way. :-)
But, yes, I agree with you that generally we should make a common
function to avoid duplicate code.
Thanks,
-- Dexuan
> From: Stephen Hemminger <[email protected]>
> Sent: Monday, December 17, 2018 10:17 AM
> To: Dexuan Cui <[email protected]>
>
> On Mon, 17 Dec 2018 18:00:29 +0000
> Dexuan Cui <[email protected]> wrote:
>
> > > From: Stephen Hemminger <[email protected]>
> > > On Thu, 13 Dec 2018 16:35:43 +0000
> > > Dexuan Cui <[email protected]> wrote:
> > >
> > > > Before 98f4c651762c, we returned zeros for unopened channels.
> > > > With 98f4c651762c, we started to return random on-stack values.
> > > >
> > > > We'd better return -EINVAL instead.
> > >
> > > The concept looks fine, but maybe it would be simpler to move it into
> > > hv_ringbuffer_get_debuginfo and have it return an error code.
> > >
> > > Since so much of the code is repeated, I would probably make a
> > > macro which generates the code as well.
> > >
> > > Something like this:
> >
> > Thanks, Stephen! Now the patch has been in char-misc's char-misc-linus
> > branch, so IMO we may as well leave it as is (considering the code here is
> > unlikely to be frqeuencly changed), and we have a smaller patch this way. :-)
> >
> > But, yes, I agree with you that generally we should make a common
> > function to avoid duplicate code.
> >
> > Thanks,
> > -- Dexuan
>
> The old code was risky because it would silently return stack garbage.
> Having an error check in get_debuginfo would eliminate that.
OK, then let me make another patch based on the latest char-misc-linus.
Thanks,
-- Dexuan
On Mon, 17 Dec 2018 18:44:12 +0000
Dexuan Cui <[email protected]> wrote:
> > From: devel <[email protected]> On Behalf Of
> > Dexuan Cui
> > Sent: Monday, December 17, 2018 10:31 AM
> > > From: Stephen Hemminger <[email protected]>
> > >
> > > The old code was risky because it would silently return stack garbage.
> > > Having an error check in get_debuginfo would eliminate that.
> >
> > OK, then let me make another patch based on the latest char-misc-linus.
> >
> > -- Dexuan
>
> Hi Stephen, your patch can apply cleanly. Let me rebase your patch to
> char-misc-linus, do a test, and then post it with your Signed-off-by and mine:
> I assume you're Ok with this. Please let me know in case it's not. :-)
>
> Thanks,
> -- Dexuan
Sure.
On Thu, 13 Dec 2018 16:35:43 +0000
Dexuan Cui <[email protected]> wrote:
> Before 98f4c651762c, we returned zeros for unopened channels.
> With 98f4c651762c, we started to return random on-stack values.
>
> We'd better return -EINVAL instead.
>
> Fixes: 98f4c651762c ("hv: move ringbuffer bus attributes to dev_groups")
> Cc: [email protected]
> Cc: K. Y. Srinivasan <[email protected]>
> Cc: Haiyang Zhang <[email protected]>
> Cc: Stephen Hemminger <[email protected]>
> Signed-off-by: Dexuan Cui <[email protected]>
The concept looks fine, but maybe it would be simpler to move it into
hv_ringbuffer_get_debuginfo and have it return an error code.
Since so much of the code is repeated, I would probably make a
macro which generates the code as well.
Something like this:
From c6bbdbcde933c85098f7b3e71650a8479d52810c Mon Sep 17 00:00:00 2001
From: Stephen Hemminger <[email protected]>
Date: Mon, 17 Dec 2018 09:13:24 -0800
Subject: [PATCH] hv: vmbus: check for ring in debug info
---
drivers/hv/ring_buffer.c | 31 +++++++++---------
drivers/hv/vmbus_drv.c | 71 ++++++++++++++++++++++++++++++++++------
include/linux/hyperv.h | 5 +--
3 files changed, 79 insertions(+), 28 deletions(-)
diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 64d0c85d5161..1f1a55e07733 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -164,26 +164,25 @@ hv_get_ringbuffer_availbytes(const struct hv_ring_buffer_info *rbi,
}
/* Get various debug metrics for the specified ring buffer. */
-void hv_ringbuffer_get_debuginfo(const struct hv_ring_buffer_info *ring_info,
- struct hv_ring_buffer_debug_info *debug_info)
+int hv_ringbuffer_get_debuginfo(const struct hv_ring_buffer_info *ring_info,
+ struct hv_ring_buffer_debug_info *debug_info)
{
u32 bytes_avail_towrite;
u32 bytes_avail_toread;
- if (ring_info->ring_buffer) {
- hv_get_ringbuffer_availbytes(ring_info,
- &bytes_avail_toread,
- &bytes_avail_towrite);
-
- debug_info->bytes_avail_toread = bytes_avail_toread;
- debug_info->bytes_avail_towrite = bytes_avail_towrite;
- debug_info->current_read_index =
- ring_info->ring_buffer->read_index;
- debug_info->current_write_index =
- ring_info->ring_buffer->write_index;
- debug_info->current_interrupt_mask =
- ring_info->ring_buffer->interrupt_mask;
- }
+ if (!ring_info->ring_buffer)
+ return -EINVAL;
+
+ hv_get_ringbuffer_availbytes(ring_info,
+ &bytes_avail_toread,
+ &bytes_avail_towrite);
+ debug_info->bytes_avail_toread = bytes_avail_toread;
+ debug_info->bytes_avail_towrite = bytes_avail_towrite;
+ debug_info->current_read_index = ring_info->ring_buffer->read_index;
+ debug_info->current_write_index = ring_info->ring_buffer->write_index;
+ debug_info->current_interrupt_mask
+ = ring_info->ring_buffer->interrupt_mask;
+ return 0;
}
EXPORT_SYMBOL_GPL(hv_ringbuffer_get_debuginfo);
diff --git a/drivers/hv/vmbus_drv.c b/drivers/hv/vmbus_drv.c
index 283d184280af..403fee01572c 100644
--- a/drivers/hv/vmbus_drv.c
+++ b/drivers/hv/vmbus_drv.c
@@ -313,10 +313,16 @@ static ssize_t out_intr_mask_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info outbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound,
+ &outbound);
+ if (ret < 0)
+ return ret;
+
return sprintf(buf, "%d\n", outbound.current_interrupt_mask);
}
static DEVICE_ATTR_RO(out_intr_mask);
@@ -326,10 +332,15 @@ static ssize_t out_read_index_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info outbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound,
+ &outbound);
+ if (ret < 0)
+ return ret;
return sprintf(buf, "%d\n", outbound.current_read_index);
}
static DEVICE_ATTR_RO(out_read_index);
@@ -340,10 +351,15 @@ static ssize_t out_write_index_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info outbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound,
+ &outbound);
+ if (ret < 0)
+ return ret;
return sprintf(buf, "%d\n", outbound.current_write_index);
}
static DEVICE_ATTR_RO(out_write_index);
@@ -354,10 +370,15 @@ static ssize_t out_read_bytes_avail_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info outbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound,
+ &outbound);
+ if (ret < 0)
+ return ret;
return sprintf(buf, "%d\n", outbound.bytes_avail_toread);
}
static DEVICE_ATTR_RO(out_read_bytes_avail);
@@ -368,10 +389,15 @@ static ssize_t out_write_bytes_avail_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info outbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound, &outbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->outbound,
+ &outbound);
+ if (ret < 0)
+ return ret;
return sprintf(buf, "%d\n", outbound.bytes_avail_towrite);
}
static DEVICE_ATTR_RO(out_write_bytes_avail);
@@ -381,10 +407,15 @@ static ssize_t in_intr_mask_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info inbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+ if (ret < 0)
+ return ret;
+
return sprintf(buf, "%d\n", inbound.current_interrupt_mask);
}
static DEVICE_ATTR_RO(in_intr_mask);
@@ -394,10 +425,15 @@ static ssize_t in_read_index_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info inbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+ if (ret < 0)
+ return ret;
+
return sprintf(buf, "%d\n", inbound.current_read_index);
}
static DEVICE_ATTR_RO(in_read_index);
@@ -407,10 +443,15 @@ static ssize_t in_write_index_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info inbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+ if (ret < 0)
+ return ret;
+
return sprintf(buf, "%d\n", inbound.current_write_index);
}
static DEVICE_ATTR_RO(in_write_index);
@@ -421,10 +462,15 @@ static ssize_t in_read_bytes_avail_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info inbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+ if (ret < 0)
+ return ret;
+
return sprintf(buf, "%d\n", inbound.bytes_avail_toread);
}
static DEVICE_ATTR_RO(in_read_bytes_avail);
@@ -435,10 +481,15 @@ static ssize_t in_write_bytes_avail_show(struct device *dev,
{
struct hv_device *hv_dev = device_to_hv_device(dev);
struct hv_ring_buffer_debug_info inbound;
+ int ret;
if (!hv_dev->channel)
return -ENODEV;
- hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+
+ ret = hv_ringbuffer_get_debuginfo(&hv_dev->channel->inbound, &inbound);
+ if (ret < 0)
+ return ret;
+
return sprintf(buf, "%d\n", inbound.bytes_avail_towrite);
}
static DEVICE_ATTR_RO(in_write_bytes_avail);
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 14131b6fae68..ed74888087f1 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -1176,8 +1176,9 @@ struct hv_ring_buffer_debug_info {
u32 bytes_avail_towrite;
};
-void hv_ringbuffer_get_debuginfo(const struct hv_ring_buffer_info *ring_info,
- struct hv_ring_buffer_debug_info *debug_info);
+
+int hv_ringbuffer_get_debuginfo(const struct hv_ring_buffer_info *ring_info,
+ struct hv_ring_buffer_debug_info *debug_info);
/* Vmbus interface */
#define vmbus_driver_register(driver) \
--
2.19.2
On Mon, 17 Dec 2018 18:00:29 +0000
Dexuan Cui <[email protected]> wrote:
> > From: Stephen Hemminger <[email protected]>
> > On Thu, 13 Dec 2018 16:35:43 +0000
> > Dexuan Cui <[email protected]> wrote:
> >
> > > Before 98f4c651762c, we returned zeros for unopened channels.
> > > With 98f4c651762c, we started to return random on-stack values.
> > >
> > > We'd better return -EINVAL instead.
> >
> > The concept looks fine, but maybe it would be simpler to move it into
> > hv_ringbuffer_get_debuginfo and have it return an error code.
> >
> > Since so much of the code is repeated, I would probably make a
> > macro which generates the code as well.
> >
> > Something like this:
>
> Thanks, Stephen! Now the patch has been in char-misc's char-misc-linus
> branch, so IMO we may as well leave it as is (considering the code here is
> unlikely to be frqeuencly changed), and we have a smaller patch this way. :-)
>
> But, yes, I agree with you that generally we should make a common
> function to avoid duplicate code.
>
> Thanks,
> -- Dexuan
The old code was risky because it would silently return stack garbage.
Having an error check in get_debuginfo would eliminate that.
> From: devel <[email protected]> On Behalf Of
> Dexuan Cui
> Sent: Monday, December 17, 2018 10:31 AM
> > From: Stephen Hemminger <[email protected]>
> >
> > The old code was risky because it would silently return stack garbage.
> > Having an error check in get_debuginfo would eliminate that.
>
> OK, then let me make another patch based on the latest char-misc-linus.
>
> -- Dexuan
Hi Stephen, your patch can apply cleanly. Let me rebase your patch to
char-misc-linus, do a test, and then post it with your Signed-off-by and mine:
I assume you're Ok with this. Please let me know in case it's not. :-)
Thanks,
-- Dexuan