Message-Id: <083251112829de95609c2220dde2e381940184cb.camel@au1.ibm.com>
Date:   Thu, 05 Mar 2020 14:38:49 +1100
From:   "Alastair D'Silva" <alastair@....ibm.com>
To:     Frederic Barrat <fbarrat@...ux.ibm.com>
Cc:     "Aneesh Kumar K . V" <aneesh.kumar@...ux.ibm.com>,
        "Oliver O'Halloran" <oohall@...il.com>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Michael Ellerman <mpe@...erman.id.au>,
        Andrew Donnellan <ajd@...ux.ibm.com>,
        Arnd Bergmann <arnd@...db.de>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Dan Williams <dan.j.williams@...el.com>,
        Vishal Verma <vishal.l.verma@...el.com>,
        Dave Jiang <dave.jiang@...el.com>,
        Ira Weiny <ira.weiny@...el.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Mauro Carvalho Chehab <mchehab+samsung@...nel.org>,
        "David S. Miller" <davem@...emloft.net>,
        Rob Herring <robh@...nel.org>,
        Anton Blanchard <anton@...abs.org>,
        Krzysztof Kozlowski <krzk@...nel.org>,
        Mahesh Salgaonkar <mahesh@...ux.vnet.ibm.com>,
        Madhavan Srinivasan <maddy@...ux.vnet.ibm.com>,
        Cédric Le Goater <clg@...d.org>,
        Anju T Sudhakar <anju@...ux.vnet.ibm.com>,
        Hari Bathini <hbathini@...ux.ibm.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Greg Kurz <groug@...d.org>,
        Nicholas Piggin <npiggin@...il.com>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>,
        Alexey Kardashevskiy <aik@...abs.ru>,
        linux-kernel@...r.kernel.org, linuxppc-dev@...ts.ozlabs.org,
        linux-nvdimm@...ts.01.org, linux-mm@...ck.org
Subject: Re: [PATCH v3 16/27] powerpc/powernv/pmem: Register a character
 device for userspace to interact with

On Tue, 2020-03-03 at 10:28 +0100, Frederic Barrat wrote:
> 
> Le 21/02/2020 à 04:27, Alastair D'Silva a écrit :
> > From: Alastair D'Silva <alastair@...ilva.org>
> > 
> > This patch introduces a character device (/dev/ocxl-scmX) which
> > further
> > patches will use to interact with userspace.
> > 
> > Signed-off-by: Alastair D'Silva <alastair@...ilva.org>
> > ---
> >   arch/powerpc/platforms/powernv/pmem/ocxl.c    | 116
> > +++++++++++++++++-
> >   .../platforms/powernv/pmem/ocxl_internal.h    |   2 +
> >   2 files changed, 116 insertions(+), 2 deletions(-)
> > 
> > diff --git a/arch/powerpc/platforms/powernv/pmem/ocxl.c
> > b/arch/powerpc/platforms/powernv/pmem/ocxl.c
> > index b8bd7e703b19..63109a870d2c 100644
> > --- a/arch/powerpc/platforms/powernv/pmem/ocxl.c
> > +++ b/arch/powerpc/platforms/powernv/pmem/ocxl.c
> > @@ -10,6 +10,7 @@
> >   #include <misc/ocxl.h>
> >   #include <linux/delay.h>
> >   #include <linux/ndctl.h>
> > +#include <linux/fs.h>
> >   #include <linux/mm_types.h>
> >   #include <linux/memory_hotplug.h>
> >   #include "ocxl_internal.h"
> > @@ -339,6 +340,9 @@ static void free_ocxlpmem(struct ocxlpmem
> > *ocxlpmem)
> >   
> >   	free_minor(ocxlpmem);
> >   
> > +	if (ocxlpmem->cdev.owner)
> > +		cdev_del(&ocxlpmem->cdev);
> > +
> >   	if (ocxlpmem->metadata_addr)
> >   		devm_memunmap(&ocxlpmem->dev, ocxlpmem->metadata_addr);
> >   
> > @@ -396,6 +400,70 @@ static int ocxlpmem_register(struct ocxlpmem
> > *ocxlpmem)
> >   	return device_register(&ocxlpmem->dev);
> >   }
> >   
> > +static void ocxlpmem_put(struct ocxlpmem *ocxlpmem)
> > +{
> > +	put_device(&ocxlpmem->dev);
> > +}
> > +
> > +static struct ocxlpmem *ocxlpmem_get(struct ocxlpmem *ocxlpmem)
> > +{
> > +	return (get_device(&ocxlpmem->dev) == NULL) ? NULL : ocxlpmem;
> > +}
> > +
> > +static struct ocxlpmem *find_and_get_ocxlpmem(dev_t devno)
> > +{
> > +	struct ocxlpmem *ocxlpmem;
> > +	int minor = MINOR(devno);
> > +	/*
> > +	 * We don't declare an RCU critical section here, as our AFU
> > +	 * is protected by a reference counter on the device. By the
> > time the
> > +	 * minor number of a device is removed from the idr, the ref
> > count of
> > +	 * the device is already at 0, so no user API will access that
> > AFU and
> > +	 * this function can't return it.
> > +	 */
> 
> I fixed something related in the ocxl driver (which had enough
> changes 
> with the introduction of the "info" device to make a similar comment 
> become wrong). See commit a58d37bce0d21. The issue is handling a 
> simultaneous open() and removal of the device through /sysfs as best
> we can.
> 
> We are on a file open path and it's not like we're going to have a 
> thousand clients, so performance is not that critical. We can take
> the 
> mutex before searching in the IDR and release it after we increment
> the 
> reference count on the device.
> But that's not enough: we could still find the device in the IDR
> while 
> it is being removed in free_ocxlpmem(). I believe the only safe way
> to 
> address it is by removing the user-facing APIs (the char device)
> before 
> calling device_unregister(). So that it's not possible to find the 
> device in file_open() if it's in the middle of being removed.
> 
>    Fred
> 
> 

Ok, I'll replicate that patch & follow your advice.

> > +	ocxlpmem = idr_find(&minors_idr, minor);
> > +	if (ocxlpmem)
> > +		ocxlpmem_get(ocxlpmem);
> > +	return ocxlpmem;
> > +}
> > +
> > +static int file_open(struct inode *inode, struct file *file)
> > +{
> > +	struct ocxlpmem *ocxlpmem;
> > +
> > +	ocxlpmem = find_and_get_ocxlpmem(inode->i_rdev);
> > +	if (!ocxlpmem)
> > +		return -ENODEV;
> > +
> > +	file->private_data = ocxlpmem;
> > +	return 0;
> > +}
> > +
> > +static int file_release(struct inode *inode, struct file *file)
> > +{
> > +	struct ocxlpmem *ocxlpmem = file->private_data;
> > +
> > +	ocxlpmem_put(ocxlpmem);
> > +	return 0;
> > +}
> > +
> > +static const struct file_operations fops = {
> > +	.owner		= THIS_MODULE,
> > +	.open		= file_open,
> > +	.release	= file_release,
> > +};
> > +
> > +/**
> > + * create_cdev() - Create the chardev in /dev for the device
> > + * @ocxlpmem: the SCM metadata
> > + * Return: 0 on success, negative on failure
> > + */
> > +static int create_cdev(struct ocxlpmem *ocxlpmem)
> > +{
> > +	cdev_init(&ocxlpmem->cdev, &fops);
> > +	return cdev_add(&ocxlpmem->cdev, ocxlpmem->dev.devt, 1);
> > +}
> > +
> >   /**
> >    * ocxlpmem_remove() - Free an OpenCAPI persistent memory device
> >    * @pdev: the PCI device information struct
> > @@ -572,6 +640,11 @@ static int probe(struct pci_dev *pdev, const
> > struct pci_device_id *ent)
> >   		goto err;
> >   	}
> >   
> > +	if (create_cdev(ocxlpmem)) {
> > +		dev_err(&pdev->dev, "Could not create character
> > device\n");
> > +		goto err;
> > +	}
> 
> As already mentioned in a previous patch, we branch to the err label
> so 
> rc needs to be set to a valid error.
> 

Ok

> 
> 
> > +
> >   	elapsed = 0;
> >   	timeout = ocxlpmem->readiness_timeout + ocxlpmem-
> > >memory_available_timeout;
> >   	while (!is_usable(ocxlpmem, false)) {
> > @@ -613,20 +686,59 @@ static struct pci_driver pci_driver = {
> >   	.shutdown = ocxlpmem_remove,
> >   };
> >   
> > +static int file_init(void)
> > +{
> > +	int rc;
> > +
> > +	mutex_init(&minors_idr_lock);
> > +	idr_init(&minors_idr);
> > +
> > +	rc = alloc_chrdev_region(&ocxlpmem_dev, 0, NUM_MINORS, "ocxl-
> > pmem");
> > +	if (rc) {
> > +		idr_destroy(&minors_idr);
> > +		pr_err("Unable to allocate OpenCAPI persistent memory
> > major number: %d\n", rc);
> > +		return rc;
> > +	}
> > +
> > +	ocxlpmem_class = class_create(THIS_MODULE, "ocxl-pmem");
> > +	if (IS_ERR(ocxlpmem_class)) {
> > +		idr_destroy(&minors_idr);
> > +		pr_err("Unable to create ocxl-pmem class\n");
> > +		unregister_chrdev_region(ocxlpmem_dev, NUM_MINORS);
> > +		return PTR_ERR(ocxlpmem_class);
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +static void file_exit(void)
> > +{
> > +	class_destroy(ocxlpmem_class);
> > +	unregister_chrdev_region(ocxlpmem_dev, NUM_MINORS);
> > +	idr_destroy(&minors_idr);
> > +}
> > +
> >   static int __init ocxlpmem_init(void)
> >   {
> > -	int rc = 0;
> > +	int rc;
> >   
> > -	rc = pci_register_driver(&pci_driver);
> > +	rc = file_init();
> >   	if (rc)
> >   		return rc;
> >   
> > +	rc = pci_register_driver(&pci_driver);
> > +	if (rc) {
> > +		file_exit();
> > +		return rc;
> > +	}
> > +
> >   	return 0;
> >   }
> >   
> >   static void ocxlpmem_exit(void)
> >   {
> >   	pci_unregister_driver(&pci_driver);
> > +	file_exit();
> >   }
> >   
> >   module_init(ocxlpmem_init);
> > diff --git a/arch/powerpc/platforms/powernv/pmem/ocxl_internal.h
> > b/arch/powerpc/platforms/powernv/pmem/ocxl_internal.h
> > index 28e2020f6355..d2d81fec7bb1 100644
> > --- a/arch/powerpc/platforms/powernv/pmem/ocxl_internal.h
> > +++ b/arch/powerpc/platforms/powernv/pmem/ocxl_internal.h
> > @@ -2,6 +2,7 @@
> >   // Copyright 2019 IBM Corp.
> >   
> >   #include <linux/pci.h>
> > +#include <linux/cdev.h>
> >   #include <misc/ocxl.h>
> >   #include <linux/libnvdimm.h>
> >   #include <linux/mm.h>
> > @@ -99,6 +100,7 @@ struct ocxlpmem_function0 {
> >   struct ocxlpmem {
> >   	struct device dev;
> >   	struct pci_dev *pdev;
> > +	struct cdev cdev;
> >   	struct ocxl_fn *ocxl_fn;
> >   	struct nd_interleave_set nd_set;
> >   	struct nvdimm_bus_descriptor bus_desc;
> > 
-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

Powered by blists - more mailing lists