Message-ID: <02e7047071f0b54b046ac472adeeb3fafabc643c.camel@redhat.com>
Date:   Fri, 11 Dec 2020 09:11:21 -0500
From:   Qian Cai <qcai@...hat.com>
To:     Saravana Kannan <saravanak@...gle.com>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Len Brown <lenb@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Ard Biesheuvel <ardb@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        Frank Rowand <frowand.list@...il.com>,
        Marc Zyngier <maz@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>
Cc:     Tomi Valkeinen <tomi.valkeinen@...com>,
        Laurent Pinchart <laurent.pinchart@...asonboard.com>,
        Grygorii Strashko <grygorii.strashko@...com>,
        kernel-team@...roid.com, linux-acpi@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-efi@...r.kernel.org,
        devicetree@...r.kernel.org,
        Stephen Rothwell <sfr@...b.auug.org.au>,
        Linux Next Mailing List <linux-next@...r.kernel.org>,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH v2 16/17] driver core: Refactor fw_devlink feature

On Fri, 2020-11-20 at 18:02 -0800, Saravana Kannan wrote:
> The current implementation of fw_devlink is very inefficient because it
> tries to get away without creating fwnode links in the name of saving
> memory usage. Past attempts to optimize runtime at the cost of memory
> usage were blocked with request for data showing that the optimization
> made significant improvement for real world scenarios.
> 
> We have those scenarios now. There have been several reports of boot
> time increase in the order of seconds in this thread [1]. Several OEMs
> and SoC manufacturers have also privately reported significant
> (350-400ms) increase in boot time due to all the parsing done by
> fw_devlink.
> 
> So this patch uses all the setup done by the previous patches in this
> series to refactor fw_devlink to be more efficient. Most of the code has
> been moved out of firmware specific (DT mostly) code into driver core.
> 
> This brings the following benefits:
> - Instead of parsing the device tree multiple times during bootup,
>   fw_devlink parses each fwnode node/property only once and creates
>   fwnode links. The rest of the fw_devlink code then just looks at these
>   fwnode links to do rest of the work.
> 
> - Makes it much easier to debug probe issue due to fw_devlink in the
>   future. fw_devlink=on blocks the probing of devices if they depend on
>   a device that hasn't been added yet. With this refactor, it'll be very
>   easy to tell what that device is because we now have a reference to
>   the fwnode of the device.
> 
> - Much easier to add fw_devlink support to ACPI and other firmware
>   types. A refactor to move the common bits from DT specific code to
>   driver core was in my TODO list as a prerequisite to adding ACPI
>   support to fw_devlink. This series gets that done.
> 
> [1] - https://lore.kernel.org/linux-omap/ea02f57e-871d-cd16-4418-c1da4bbc4696@ti.com/
> Signed-off-by: Saravana Kannan <saravanak@...gle.com>
> Tested-by: Laurent Pinchart <laurent.pinchart@...asonboard.com>
> Tested-by: Grygorii Strashko <grygorii.strashko@...com>

Reverting this commit and its dependency:

2d09e6eb4a6f driver core: Delete pointless parameter in fwnode_operations.add_links

from today's linux-next fixed a boot crash on an arm64 Thunder X2 server.

.config: https://cailca.coding.net/public/linux/mm/git/files/master/arm64.config

[   57.413929][    T1] ACPI: 5 ACPI AML tables successfully acquired and loaded
[   60.571643][    T1] ACPI: Interpreter enabled
[   60.576104][    T1] ACPI: Using GIC for interrupt routing
[   60.582474][    T1] ACPI: MCFG table detected, 1 entries
[   60.588051][    T1] ACPI: IORT: SMMU-v3[402300000] Mapped to Proximity domain 0
[   60.601374][    T1] Unable to handle kernel paging request at virtual address dfff800000000000
[   60.610146][    T1] Mem abort info:
[   60.613694][    T1]   ESR = 0x96000004
[   60.617496][    T1]   EC = 0x25: DABT (current EL), IL = 32 bits
[   60.623616][    T1]   SET = 0, FnV = 0
[   60.627420][    T1]   EA = 0, S1PTW = 0
[   60.631304][    T1] Data abort info:
[   60.634957][    T1]   ISV = 0, ISS = 0x00000004
[   60.639546][    T1]   CM = 0, WnR = 0
[   60.643255][    T1] [dfff800000000000] address between user and kernel address ranges
[   60.651226][    T1] Internal error: Oops: 96000004 [#1] SMP
[   60.656864][    T1] Modules linked in:
[   60.660658][    T1] CPU: 38 PID: 1 Comm: swapper/0 Tainted: G        W         5.10.0-rc7-next-20201211 #2
[   60.670424][    T1] Hardware name: HPE Apollo 70             /C01_APACHE_MB         , BIOS L50_5.13_1.16 07/29/2020
[   60.680979][    T1] pstate: 10400009 (nzcV daif +PAN -UAO -TCO BTYPE=--)
[   60.687757][    T1] pc : device_add+0xf60/0x16b0
[   60.692430][    T1] lr : device_add+0xf08/0x16b0
[   60.697098][    T1] sp : ffff0000063bf7d0
[   60.701147][    T1] x29: ffff0000063bf7d0 x28: ffff00001f760810 
[   60.707226][    T1] x27: fffffffffffffff8 x26: ffff00001f760858 
[   60.713304][    T1] x25: ffff00001f760c58 x24: fffffffffffffff8 
[   60.719381][    T1] x23: 1fffe00003eec10b x22: ffff8000190d0260 
[   60.725458][    T1] x21: ffff800011dba708 x20: 0000000000000000 
[   60.731535][    T1] x19: 1fffe00000c77f10 x18: 1fffe001cf0d53ed 
[   60.737616][    T1] x17: 0000000000000000 x16: 1ffff000033676f1 
[   60.743709][    T1] x15: 0000000000000000 x14: ffff800011731e34 
[   60.749786][    T1] x13: ffff700003217fb9 x12: 1ffff00003217fb8 
[   60.755864][    T1] x11: 1ffff00003217fb8 x10: ffff700003217fb8 
[   60.761940][    T1] x9 : dfff800000000000 x8 : ffff8000190bfdc7 
[   60.768017][    T1] x7 : 0000000000000001 x6 : ffff700003217fb9 
[   60.774094][    T1] x5 : ffff700003217fb9 x4 : ffff000006324a80 
[   60.780170][    T1] x3 : 1fffe00000c64951 x2 : 0000000000000000 
[   60.786247][    T1] x1 : 0000000000000000 x0 : dfff800000000000 
[   60.792324][    T1] Call trace:
[   60.795495][    T1]  device_add+0xf60/0x16b0
__fw_devlink_link_to_consumers at drivers/base/core.c:1583
(inlined by) fw_devlink_link_device at drivers/base/core.c:1726
(inlined by) device_add at drivers/base/core.c:3088
[   60.799813][    T1]  platform_device_add+0x274/0x628
[   60.804833][    T1]  acpi_iort_init+0x9d8/0xc50
[   60.809415][    T1]  acpi_init+0x45c/0x4e8
[   60.813556][    T1]  do_one_initcall+0x170/0xb70
[   60.818224][    T1]  kernel_init_freeable+0x6a8/0x734
[   60.823332][    T1]  kernel_init+0x18/0x12c
[   60.827566][    T1]  ret_from_fork+0x10/0x1c
[   60.831890][    T1] Code: f2fbffe0 d100205b d343fc41 aa1b03f8 (38e06820) 
[   60.838756][    T1] ---[ end trace fa5c8ce17a226d83 ]---
[   60.844127][    T1] Kernel panic - not syncing: Oops: Fatal exception
[   60.850881][    T1] SMP: stopping secondary CPUs
[   60.855733][    T1] ---[ end Kernel panic - not syncing: Oops: Fatal exception ]---
> ---
>  drivers/base/core.c    | 325 ++++++++++++++++++++++++++++++-----------
>  include/linux/device.h |   5 -
>  2 files changed, 238 insertions(+), 92 deletions(-)
> 
> diff --git a/drivers/base/core.c b/drivers/base/core.c
> index 1873cecb0cc4..9edf9084fc98 100644
> --- a/drivers/base/core.c
> +++ b/drivers/base/core.c
> @@ -46,8 +46,6 @@ early_param("sysfs.deprecated", sysfs_deprecated_setup);
>  #endif
>  
>  /* Device links support. */
> -static LIST_HEAD(wait_for_suppliers);
> -static DEFINE_MUTEX(wfs_lock);
>  static LIST_HEAD(deferred_sync);
>  static unsigned int defer_sync_state_count = 1;
>  static DEFINE_MUTEX(fwnode_link_lock);
> @@ -803,74 +801,6 @@ struct device_link *device_link_add(struct device *consumer,
>  }
>  EXPORT_SYMBOL_GPL(device_link_add);
>  
> -/**
> - * device_link_wait_for_supplier - Add device to wait_for_suppliers list
> - * @consumer: Consumer device
> - *
> - * Marks the @consumer device as waiting for suppliers to become available by
> - * adding it to the wait_for_suppliers list. The consumer device will never be
> - * probed until it's removed from the wait_for_suppliers list.
> - *
> - * The caller is responsible for adding the links to the supplier devices once
> - * they are available and removing the @consumer device from the
> - * wait_for_suppliers list once links to all the suppliers have been created.
> - *
> - * This function is NOT meant to be called from the probe function of the
> - * consumer but rather from code that creates/adds the consumer device.
> - */
> -static void device_link_wait_for_supplier(struct device *consumer,
> -					  bool need_for_probe)
> -{
> -	mutex_lock(&wfs_lock);
> -	list_add_tail(&consumer->links.needs_suppliers, &wait_for_suppliers);
> -	consumer->links.need_for_probe = need_for_probe;
> -	mutex_unlock(&wfs_lock);
> -}
> -
> -static void device_link_wait_for_mandatory_supplier(struct device *consumer)
> -{
> -	device_link_wait_for_supplier(consumer, true);
> -}
> -
> -static void device_link_wait_for_optional_supplier(struct device *consumer)
> -{
> -	device_link_wait_for_supplier(consumer, false);
> -}
> -
> -/**
> - * device_link_add_missing_supplier_links - Add links from consumer devices to
> - *					    supplier devices, leaving any
> - *					    consumer with inactive suppliers on
> - *					    the wait_for_suppliers list
> - *
> - * Loops through all consumers waiting on suppliers and tries to add all their
> - * supplier links. If that succeeds, the consumer device is removed from
> - * wait_for_suppliers list. Otherwise, they are left in the wait_for_suppliers
> - * list.  Devices left on the wait_for_suppliers list will not be probed.
> - *
> - * The fwnode add_links callback is expected to return 0 if it has found and
> - * added all the supplier links for the consumer device. It should return an
> - * error if it isn't able to do so.
> - *
> - * The caller of device_link_wait_for_supplier() is expected to call this once
> - * it's aware of potential suppliers becoming available.
> - */
> -static void device_link_add_missing_supplier_links(void)
> -{
> -	struct device *dev, *tmp;
> -
> -	mutex_lock(&wfs_lock);
> -	list_for_each_entry_safe(dev, tmp, &wait_for_suppliers,
> -				 links.needs_suppliers) {
> -		int ret = fwnode_call_int_op(dev->fwnode, add_links, dev);
> -		if (!ret)
> -			list_del_init(&dev->links.needs_suppliers);
> -		else if (ret != -ENODEV)
> -			dev->links.need_for_probe = false;
> -	}
> -	mutex_unlock(&wfs_lock);
> -}
> -
>  #ifdef CONFIG_SRCU
>  static void __device_link_del(struct kref *kref)
>  {
> @@ -1195,9 +1125,8 @@ void device_links_driver_bound(struct device *dev)
>  	 * the device links it needs to or make new device links as it needs
>  	 * them. So, it no longer needs to wait on any suppliers.
>  	 */
> -	mutex_lock(&wfs_lock);
> -	list_del_init(&dev->links.needs_suppliers);
> -	mutex_unlock(&wfs_lock);
> +	if (dev->fwnode && dev->fwnode->dev == dev)
> +		fwnode_links_purge_suppliers(dev->fwnode);
>  	device_remove_file(dev, &dev_attr_waiting_for_supplier);
>  
>  	device_links_write_lock();
> @@ -1488,10 +1417,6 @@ static void device_links_purge(struct device *dev)
>  	if (dev->class == &devlink_class)
>  		return;
>  
> -	mutex_lock(&wfs_lock);
> -	list_del(&dev->links.needs_suppliers);
> -	mutex_unlock(&wfs_lock);
> -
>  	/*
>  	 * Delete all of the remaining links from this device to any other
>  	 * devices (either consumers or suppliers).
> @@ -1561,19 +1486,246 @@ static void fw_devlink_parse_fwtree(struct fwnode_handle *fwnode)
>  		fw_devlink_parse_fwtree(child);
>  }
>  
> -static void fw_devlink_link_device(struct device *dev)
> +/**
> + * fw_devlink_create_devlink - Create a device link from a consumer to fwnode
> + * @con - Consumer device for the device link
> + * @sup_handle - fwnode handle of supplier
> + *
> + * This function will try to create a device link between the consumer device
> + * @con and the supplier device represented by @sup_handle.
> + *
> + * The supplier has to be provided as a fwnode because incorrect cycles in
> + * fwnode links can sometimes cause the supplier device to never be created.
> + * This function detects such cases and returns an error if it cannot create a
> + * device link from the consumer to a missing supplier.
> + *
> + * Returns,
> + * 0 on successfully creating a device link
> + * -EINVAL if the device link cannot be created as expected
> + * -EAGAIN if the device link cannot be created right now, but it may be
> + *  possible to do that in the future
> + */
> +static int fw_devlink_create_devlink(struct device *con,
> +				     struct fwnode_handle *sup_handle, u32 flags)
> +{
> +	struct device *sup_dev;
> +	int ret = 0;
> +
> +	sup_dev = get_dev_from_fwnode(sup_handle);
> +	if (sup_dev) {
> +		/*
> +		 * If this fails, it is due to cycles in device links.  Just
> +		 * give up on this link and treat it as invalid.
> +		 */
> +		if (!device_link_add(con, sup_dev, flags))
> +			ret = -EINVAL;
> +
> +		goto out;
> +	}
> +
> +	/*
> +	 * DL_FLAG_SYNC_STATE_ONLY doesn't block probing and supports
> +	 * cycles. So cycle detection isn't necessary and shouldn't be
> +	 * done.
> +	 */
> +	if (flags & DL_FLAG_SYNC_STATE_ONLY)
> +		return -EAGAIN;
> +
> +	/*
> +	 * If we can't find the supplier device from its fwnode, it might be
> +	 * due to a cyclic dependency between fwnodes. Some of these cycles can
> +	 * be broken by applying logic. Check for these types of cycles and
> +	 * break them so that devices in the cycle probe properly.
> +	 *
> +	 * If the supplier's parent is dependent on the consumer, then
> +	 * the consumer-supplier dependency is a false dependency. So,
> +	 * treat it as an invalid link.
> +	 */
> +	sup_dev = fwnode_get_next_parent_dev(sup_handle);
> +	if (sup_dev && device_is_dependent(con, sup_dev)) {
> +		dev_dbg(con, "Not linking to %pfwP - False link\n",
> +			sup_handle);
> +		ret = -EINVAL;
> +	} else {
> +		/*
> +		 * Can't check for cycles or no cycles. So let's try
> +		 * again later.
> +		 */
> +		ret = -EAGAIN;
> +	}
> +
> +out:
> +	put_device(sup_dev);
> +	return ret;
> +}
> +
> +/**
> + * __fw_devlink_link_to_consumers - Create device links to consumers of a device
> + * @dev - Device that needs to be linked to its consumers
> + *
> + * This function looks at all the consumer fwnodes of @dev and creates device
> + * links between the consumer device and @dev (supplier).
> + *
> + * If the consumer device has not been added yet, then this function creates a
> + * SYNC_STATE_ONLY link between @dev (supplier) and the closest ancestor device
> + * of the consumer fwnode. This is necessary to make sure @dev doesn't get a
> + * sync_state() callback before the real consumer device gets to be added and
> + * then probed.
> + *
> + * Once device links are created from the real consumer to @dev (supplier), the
> + * fwnode links are deleted.
> + */
> +static void __fw_devlink_link_to_consumers(struct device *dev)
> +{
> +	struct fwnode_handle *fwnode = dev->fwnode;
> +	struct fwnode_link *link, *tmp;
> +
> +	list_for_each_entry_safe(link, tmp, &fwnode->consumers, s_hook) {
> +		u32 dl_flags = fw_devlink_get_flags();
> +		struct device *con_dev;
> +		bool own_link = true;
> +		int ret;
> +
> +		con_dev = get_dev_from_fwnode(link->consumer);
> +		/*
> +		 * If consumer device is not available yet, make a "proxy"
> +		 * SYNC_STATE_ONLY link from the consumer's parent device to
> +		 * the supplier device. This is necessary to make sure the
> +		 * supplier doesn't get a sync_state() callback before the real
> +		 * consumer can create a device link to the supplier.
> +		 *
> +		 * This proxy link step is needed to handle the case where the
> +		 * consumer's parent device is added before the supplier.
> +		 */
> +		if (!con_dev) {
> +			con_dev = fwnode_get_next_parent_dev(link->consumer);
> +			/*
> +			 * However, if the consumer's parent device is also the
> +			 * parent of the supplier, don't create a
> +			 * consumer-supplier link from the parent to its child
> +			 * device. Such a dependency is impossible.
> +			 */
> +			if (con_dev &&
> +			    fwnode_is_ancestor_of(con_dev->fwnode, fwnode)) {
> +				put_device(con_dev);
> +				con_dev = NULL;
> +			} else {
> +				own_link = false;
> +				dl_flags = DL_FLAG_SYNC_STATE_ONLY;
> +			}
> +		}
> +
> +		if (!con_dev)
> +			continue;
> +
> +		ret = fw_devlink_create_devlink(con_dev, fwnode, dl_flags);
> +		put_device(con_dev);
> +		if (!own_link || ret == -EAGAIN)
> +			continue;
> +
> +		list_del(&link->s_hook);
> +		list_del(&link->c_hook);
> +		kfree(link);
> +	}
> +}
> +
> +/**
> + * __fw_devlink_link_to_suppliers - Create device links to suppliers of a device
> + * @dev - The consumer device that needs to be linked to its suppliers
> + * @fwnode - Root of the fwnode tree that is used to create device links
> + *
> + * This function looks at all the supplier fwnodes of fwnode tree rooted at
> + * @fwnode and creates device links between @dev (consumer) and all the
> + * supplier devices of the entire fwnode tree at @fwnode.
> + *
> + * The function creates normal (non-SYNC_STATE_ONLY) device links between @dev
> + * and the real suppliers of @dev. Once these device links are created, the
> + * fwnode links are deleted. When such device links are successfully created,
> + * this function is called recursively on those supplier devices. This is
> + * needed to detect and break some invalid cycles in fwnode links.  See
> + * fw_devlink_create_devlink() for more details.
> + *
> + * In addition, it also looks at all the suppliers of the entire fwnode tree
> + * because some of the child devices of @dev that have not been added yet
> + * (because @dev hasn't probed) might already have their suppliers added to
> + * driver core. So, this function creates SYNC_STATE_ONLY device links between
> + * @dev (consumer) and these suppliers to make sure they don't execute their
> + * sync_state() callbacks before these child devices have a chance to create
> + * their device links. The fwnode links that correspond to the child devices
> + * aren't delete because they are needed later to create the device links
> + * between the real consumer and supplier devices.
> + */
> +static void __fw_devlink_link_to_suppliers(struct device *dev,
> +					   struct fwnode_handle *fwnode)
>  {
> -	int fw_ret;
> +	bool own_link = (dev->fwnode == fwnode);
> +	struct fwnode_link *link, *tmp;
> +	struct fwnode_handle *child = NULL;
> +	u32 dl_flags;
> +
> +	if (own_link)
> +		dl_flags = fw_devlink_get_flags();
> +	else
> +		dl_flags = DL_FLAG_SYNC_STATE_ONLY;
>  
> -	device_link_add_missing_supplier_links();
> +	list_for_each_entry_safe(link, tmp, &fwnode->suppliers, c_hook) {
> +		int ret;
> +		struct device *sup_dev;
> +		struct fwnode_handle *sup = link->supplier;
> +
> +		ret = fw_devlink_create_devlink(dev, sup, dl_flags);
> +		if (!own_link || ret == -EAGAIN)
> +			continue;
>  
> -	if (fw_devlink_flags && fwnode_has_op(dev->fwnode, add_links)) {
> -		fw_ret = fwnode_call_int_op(dev->fwnode, add_links, dev);
> -		if (fw_ret == -ENODEV && !fw_devlink_is_permissive())
> -			device_link_wait_for_mandatory_supplier(dev);
> -		else if (fw_ret)
> -			device_link_wait_for_optional_supplier(dev);
> +		list_del(&link->s_hook);
> +		list_del(&link->c_hook);
> +		kfree(link);
> +
> +		/* If no device link was created, nothing more to do. */
> +		if (ret)
> +			continue;
> +
> +		/*
> +		 * If a device link was successfully created to a supplier, we
> +		 * now need to try and link the supplier to all its suppliers.
> +		 *
> +		 * This is needed to detect and delete false dependencies in
> +		 * fwnode links that haven't been converted to a device link
> +		 * yet. See comments in fw_devlink_create_devlink() for more
> +		 * details on the false dependency.
> +		 *
> +		 * Without deleting these false dependencies, some devices will
> +		 * never probe because they'll keep waiting for their false
> +		 * dependency fwnode links to be converted to device links.
> +		 */
> +		sup_dev = get_dev_from_fwnode(sup);
> +		__fw_devlink_link_to_suppliers(sup_dev, sup_dev->fwnode);
> +		put_device(sup_dev);
>  	}
> +
> +	/*
> +	 * Make "proxy" SYNC_STATE_ONLY device links to represent the needs of
> +	 * all the descendants. This proxy link step is needed to handle the
> +	 * case where the supplier is added before the consumer's parent device
> +	 * (@dev).
> +	 */
> +	while ((child = fwnode_get_next_available_child_node(fwnode, child)))
> +		__fw_devlink_link_to_suppliers(dev, child);
> +}
> +
> +static void fw_devlink_link_device(struct device *dev)
> +{
> +	struct fwnode_handle *fwnode = dev->fwnode;
> +
> +	if (!fw_devlink_flags)
> +		return;
> +
> +	fw_devlink_parse_fwtree(fwnode);
> +
> +	mutex_lock(&fwnode_link_lock);
> +	__fw_devlink_link_to_consumers(dev);
> +	__fw_devlink_link_to_suppliers(dev, fwnode);
> +	mutex_unlock(&fwnode_link_lock);
>  }
>  
>  /* Device links support end. */
> @@ -2431,7 +2583,6 @@ void device_initialize(struct device *dev)
>  #endif
>  	INIT_LIST_HEAD(&dev->links.consumers);
>  	INIT_LIST_HEAD(&dev->links.suppliers);
> -	INIT_LIST_HEAD(&dev->links.needs_suppliers);
>  	INIT_LIST_HEAD(&dev->links.defer_sync);
>  	dev->links.status = DL_DEV_NO_DRIVER;
>  }
> diff --git a/include/linux/device.h b/include/linux/device.h
> index 1e771ea4dca6..89bb8b84173e 100644
> --- a/include/linux/device.h
> +++ b/include/linux/device.h
> @@ -351,18 +351,13 @@ enum dl_dev_state {
>   * struct dev_links_info - Device data related to device links.
>   * @suppliers: List of links to supplier devices.
>   * @consumers: List of links to consumer devices.
> - * @needs_suppliers: Hook to global list of devices waiting for suppliers.
>   * @defer_sync: Hook to global list of devices that have deferred sync_state.
> - * @need_for_probe: If needs_suppliers is on a list, this indicates if the
> - *		    suppliers are needed for probe or not.
>   * @status: Driver status information.
>   */
>  struct dev_links_info {
>  	struct list_head suppliers;
>  	struct list_head consumers;
> -	struct list_head needs_suppliers;
>  	struct list_head defer_sync;
> -	bool need_for_probe;
>  	enum dl_dev_state status;
>  };
>  

Powered by blists - more mailing lists