by Saravana Kannan

[permalink] [raw]

Subject: Re: [PATCH v1 12/18] driver core: Add fw_devlink_parse_fwtree()

On Mon, Nov 16, 2020 at 8:25 AM Rafael J. Wysocki <[email protected]> wrote:
>
> On Thu, Nov 5, 2020 at 12:24 AM Saravana Kannan <[email protected]> wrote:
> >
> > This function is a wrapper around fwnode_operations.add_links().
> >
> > This function parses each node in a fwnode tree and create fwnode links
> > for each of those nodes. The information for creating the fwnode links
> > (the supplier and consumer fwnode) is obtained by parsing the properties
> > in each of the fwnodes.
> >
> > This function also ensures that no fwnode is parsed more than once by
> > marking the fwnodes as parsed.
> >
> > Signed-off-by: Saravana Kannan <[email protected]>
> > ---
> > drivers/base/core.c | 19 +++++++++++++++++++
> > include/linux/fwnode.h | 3 +++
> > 2 files changed, 22 insertions(+)
> >
> > diff --git a/drivers/base/core.c b/drivers/base/core.c
> > index 4a0907574646..ee28d8c7ee85 100644
> > --- a/drivers/base/core.c
> > +++ b/drivers/base/core.c
> > @@ -1543,6 +1543,25 @@ static bool fw_devlink_is_permissive(void)
> > return fw_devlink_flags == DL_FLAG_SYNC_STATE_ONLY;
> > }
> >
> > +static void fw_devlink_parse_fwnode(struct fwnode_handle *fwnode)
> > +{
> > + if (fwnode->flags & FWNODE_FLAG_LINKS_ADDED)
> > + return;
>
> Why is the flag needed?
>
> Duplicate links won't be created anyway and it doesn't cause the tree
> walk to be terminated.

To avoid parsing a fwnode more than once. The cumulative impact of the
repeated parsing is actually quite high.

And I intentionally didn't do this check at the tree walk level
because DT overlay can add/remove/change individual fwnodes and I want
to reparse those when they are added while avoiding parsing other
nodes that have already been parsed and not changed by DT overlay.

>
> > +
> > + fwnode_call_int_op(fwnode, add_links, NULL);
> > + fwnode->flags |= FWNODE_FLAG_LINKS_ADDED;
> > +}
> > +
> > +static void fw_devlink_parse_fwtree(struct fwnode_handle *fwnode)
> > +{
> > + struct fwnode_handle *child = NULL;
> > +
> > + fw_devlink_parse_fwnode(fwnode);
> > +
> > + while ((child = fwnode_get_next_available_child_node(fwnode, child)))
>
> I'd prefer
>
> for (child = NULL; child; child =
> fwnode_get_next_available_child_node(fwnode, child))

I was about to change to this and then realized it won't work. It
would have to be

for (child = fwnode_get_next_available_child_node(fwnode, NULL));
child;
child = fwnode_get_next_available_child_node(fwnode, child))

Is that really better? The while() seems a lot more readable to me. I
don't have a strong opinion, so I'll go with whatever you say after
reading this.

>
> > + fw_devlink_parse_fwtree(child);
> > +}
> > +
> > static void fw_devlink_link_device(struct device *dev)
> > {
> > int fw_ret;
> > diff --git a/include/linux/fwnode.h b/include/linux/fwnode.h
> > index ec02e1e939cc..9aaf9e4f3994 100644
> > --- a/include/linux/fwnode.h
> > +++ b/include/linux/fwnode.h
> > @@ -15,12 +15,15 @@
> > struct fwnode_operations;
> > struct device;
> >
>
> Description here, please.

Ack

>
> > +#define FWNODE_FLAG_LINKS_ADDED BIT(0)
> > +
> > struct fwnode_handle {
> > struct fwnode_handle *secondary;
> > const struct fwnode_operations *ops;
> > struct device *dev;
> > struct list_head suppliers;
> > struct list_head consumers;
> > + u32 flags;
>
> That's a bit wasteful. Maybe u8 would suffice for the time being?

Ack.

-Saravana

2020-11-21 02:04:10

by Saravana Kannan

[permalink] [raw]

Subject: Re: [PATCH v1 14/18] driver core: Use device's fwnode to check if it is waiting for suppliers

On Mon, Nov 16, 2020 at 8:34 AM Rafael J. Wysocki <[email protected]> wrote:
>
> On Thu, Nov 5, 2020 at 12:24 AM Saravana Kannan <[email protected]> wrote:
> >
> > To check if a device is still waiting for its supplier devices to be
> > added, we used to check if the devices is in a global
> > waiting_for_suppliers list. Since the global list will be deleted in
> > subsequent patches, this patch stops using this check.
>
> My kind of educated guess is that you want to drop
> waiting_for_suppliers and that's why you want to use supplier links
> here.

Yes, and a device would never be added waiting_for_suppliers list.

> >
> > Instead, this patch uses a more device specific check. It checks if the
> > device's fwnode has any fwnode links that haven't been converted to
> > device links yet.
> >
> > Signed-off-by: Saravana Kannan <[email protected]>
> > ---
> > drivers/base/core.c | 18 ++++++++----------
> > 1 file changed, 8 insertions(+), 10 deletions(-)
> >
> > diff --git a/drivers/base/core.c b/drivers/base/core.c
> > index 4ae5f2885ac5..d51dd564add1 100644
> > --- a/drivers/base/core.c
> > +++ b/drivers/base/core.c
> > @@ -51,6 +51,7 @@ static DEFINE_MUTEX(wfs_lock);
> > static LIST_HEAD(deferred_sync);
> > static unsigned int defer_sync_state_count = 1;
> > static DEFINE_MUTEX(fwnode_link_lock);
> > +static bool fw_devlink_is_permissive(void);
> >
> > /**
> > * fwnode_link_add - Create a link between two fwnode_handles.
> > @@ -994,13 +995,13 @@ int device_links_check_suppliers(struct device *dev)
> > * Device waiting for supplier to become available is not allowed to
> > * probe.
> > */
> > - mutex_lock(&wfs_lock);
> > - if (!list_empty(&dev->links.needs_suppliers) &&
> > - dev->links.need_for_probe) {
> > - mutex_unlock(&wfs_lock);
> > + mutex_lock(&fwnode_link_lock);
> > + if (dev->fwnode && !list_empty(&dev->fwnode->suppliers) &&
> > + !fw_devlink_is_permissive()) {
> > + mutex_unlock(&fwnode_link_lock);
> > return -EPROBE_DEFER;
> > }
> > - mutex_unlock(&wfs_lock);
> > + mutex_unlock(&fwnode_link_lock);
> >
> > device_links_write_lock();
> >
> > @@ -1166,10 +1167,7 @@ static ssize_t waiting_for_supplier_show(struct device *dev,
> > bool val;
> >
> > device_lock(dev);
> > - mutex_lock(&wfs_lock);
> > - val = !list_empty(&dev->links.needs_suppliers)
> > - && dev->links.need_for_probe;
> > - mutex_unlock(&wfs_lock);
>
> Why isn't the lock needed any more?
>
> Or maybe it wasn't needed previously too?

Yeah, I sent a separate patch for dropping this lock [1]. But I didn't
want to wait for that to land to write this series. The lock wasn't
needed in the first place and it was causing a lockdep warning.

>
> > + val = !list_empty(&dev->fwnode->suppliers);
> > device_unlock(dev);
> > return sysfs_emit(buf, "%u\n", val);
> > }
> > @@ -2226,7 +2224,7 @@ static int device_add_attrs(struct device *dev)
> > goto err_remove_dev_groups;
> > }
> >
> > - if (fw_devlink_flags && !fw_devlink_is_permissive()) {
> > + if (fw_devlink_flags && !fw_devlink_is_permissive() && dev->fwnode) {
>
> And why is this change needed?

Because if a device doesn't have a fwnode, it can't ever be waiting on
a supplier. Also, the "show" function dereferences
dev->fwnode->suppliers.

-Saravana

[1] - https://lore.kernel.org/lkml/[email protected]/
Ignore the 1/2 thing. There's only 1 relevant patch.