Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 14 Jul 2022 17:43:23 -0400
From:   Phil Auld <pauld@redhat.com>
To:     Yury Norov <yury.norov@gmail.com>
Cc:     linux-kernel@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "Rafael J . Wysocki" <rafael@kernel.org>,
        Barry Song <21cnbao@gmail.com>,
        Tian Tao <tiantao6@hisilicon.com>
Subject: Re: [PATCH v4 RESEND] drivers/base: fix userspace break from using
 bin_attributes for cpumap and cpulist
Message-ID: <YtCN+wXGC5U0BR9U@lorien.usersys.redhat.com>
References: <20220714183021.2924119-1-pauld@redhat.com>
 <YtBvQUzRn5idNA5J@yury-laptop>
 <YtCFjZJFpQCgBiyF@lorien.usersys.redhat.com>
 <YtCLoMlYzjgBqhia@yury-laptop>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YtCLoMlYzjgBqhia@yury-laptop>
Precedence: bulk

On Thu, Jul 14, 2022 at 02:33:20PM -0700 Yury Norov wrote:
> On Thu, Jul 14, 2022 at 05:07:25PM -0400, Phil Auld wrote:
> > Hi Yury,
> > 
> > On Thu, Jul 14, 2022 at 12:32:17PM -0700 Yury Norov wrote:
> > > On Thu, Jul 14, 2022 at 02:30:21PM -0400, Phil Auld wrote:
> > > > Using bin_attributes with a 0 size causes fstat and friends to return that 0 size.
> > > > This breaks userspace code that retrieves the size before reading the file. Rather
> > > > than reverting 75bd50fa841 ("drivers/base/node.c: use bin_attribute to break the size
> > > > limitation of cpumap ABI") let's put in a size value at compile time. Use direct
> > > > comparison and a worst-case maximum to ensure compile time constants. For cpulist the
> > > > max is on the order of NR_CPUS * (ceil(log10(NR_CPUS)) + 1) which for 8192 is 40960
> > > > (8192 * 5). In order to get near that you'd need a system with every other CPU on one
> > > > node or something similar. e.g. (0,2,4,8, ... ).
> > > 
> > > My python says:
> > >         >>> len(str(list(range(0, 8194, 2))).replace(" ", ''))
> > >         19931
> > > 
> > > Where the list looks like:
> > >         [0,2,4,6,8,10,...,8190,8192]
> > >         
> > > So excluding open and close braces, max length of the cpu list is
> > > 19929 bytes, which is almost3 times smaller than your estimation
> > > (8192 * 7 = 57344).
> > > 
> > > 
> > > For NR_CPUS == 16x8192:
> > >         >>> len(str(list(range(0, 8194 * 16, 2))).replace(" ", '')) - 2
> > >         403308
> > >         >>> 8192 * 16 * 7
> > >         917504
> > > 
> > > For NR_CPUS == 128x8192:
> > >         >>> len(str(list(range(0, 8194 * 128, 2))).replace(" ", '')) - 2
> > >         3639774
> > >         >>> 8192 * 16 * 7
> > >         7340032
> > > 
> > > Looks like it 2x overestimates for large lists, and 4x for standard
> > > 256-bit mask.
> > 
> > Thanks! It's totally possible my math is wrong. But I'm not seeing a
> > suggestion here. What would you like the formula to be?
> > 
> > I don't have any attachment to the numbers that are in there now. I'm not
> > surprised it's 2x since it counts all of them. I could just divide it by 2
> > I suppose.
> 
> Agree.
> 
> Actually, I accidentally removed the sentence below from the previous email:
> 
> I think there's no simple formula for this, so dividing by 2 would
> probably be a simplest solution. 
>

Right! That's what I've got now :)   I'll have a new version after I get
it built and sanity tested.

Thanks again,
Phil


> > > Before, it was possible to fit ~1800 cpus into 4k page, after - 585.
> > >
> > 
> > I don't understand this part. Nothing I did changes how the files are
> > actually built I think.
> 
> I just meant that we'll significantly increase requirements for the
> size.
>  
> > > > To simplify the math and support
> > > > larger NR_CPUS in the future we are using NR_CPUS * 7. We also set it to a min of
> > > > PAGE_SIZE to retain the older behavior for smaller NR_CPUS. The cpumap file wants to
> > > > be something like NR_CPUS/4 + NR_CPUS/32, for the ","s so for simplicity we are using
> > > > NR_CPUS/2.
> > > 
> > > This again overestimates almost twice. In this case, NR_CPUS * 9/32 - 1
> > > is a precise value, if I didn't screw up. Why don't you just use it?
> > >
> > 
> > I was just keeping it simple since it's used twice. But I can do it this way. 
> > 
> > 
> > Cheers,
> > Phil
> > 
> > > > Add a set of macros for these values to cpumask.h so they can be used in multiple places.
> > > > Apply these to the handful of such files in drivers/base/topology.c as well as node.c.
> > > > 
> > > > On an 80 cpu 4-node sytem (NR_CPUS == 8192)
> > > > 
> > > > before:
> > > > 
> > > > -r--r--r--. 1 root root 0 Jul 12 14:08 /sys/devices/system/node/node0/cpulist
> > > > -r--r--r--. 1 root root 0 Jul 11 17:25 /sys/devices/system/node/node0/cpumap
> > > > 
> > > > after:
> > > > 
> > > > -r--r--r--. 1 root root 57344 Jul 13 11:32 /sys/devices/system/node/node0/cpulist
> > > > -r--r--r--. 1 root root  4096 Jul 13 11:31 /sys/devices/system/node/node0/cpumap
> > > > 
> > > > CONFIG_NR_CPUS = 16384
> > > > -r--r--r--. 1 root root 114688 Jul 13 14:03 /sys/devices/system/node/node0/cpulist
> > > > -r--r--r--. 1 root root   8192 Jul 13 14:02 /sys/devices/system/node/node0/cpumap
> > > > 
> > > > Fixes: 75bd50fa841 ("drivers/base/node.c: use bin_attribute to break the size limitation of cpumap ABI")
> > > > Fixes: bb9ec13d156 ("topology: use bin_attribute to break the size limitation of cpumap ABI")
> > > > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> > > > Cc: "Rafael J. Wysocki" <rafael@kernel.org>
> > > > Signed-off-by: Phil Auld <pauld@redhat.com>
> > > > ---
> > > > 
> > > > v2: Fix cpumap size calculation. Increase multiplier for cpulist size.
> > > > 
> > > > v3: Add comments in code.
> > > > 
> > > > v4: Define constants in cpumask.h. Move comments there. Also fix
> > > > topology.c.
> > > > 
> > > > 
> > > >  drivers/base/node.c     |  4 ++--
> > > >  drivers/base/topology.c | 32 ++++++++++++++++----------------
> > > >  include/linux/cpumask.h | 16 ++++++++++++++++
> > > >  3 files changed, 34 insertions(+), 18 deletions(-)
> > > > 
> > > > diff --git a/drivers/base/node.c b/drivers/base/node.c
> > > > index 0ac6376ef7a1..eb0f43784c2b 100644
> > > > --- a/drivers/base/node.c
> > > > +++ b/drivers/base/node.c
> > > > @@ -45,7 +45,7 @@ static inline ssize_t cpumap_read(struct file *file, struct kobject *kobj,
> > > >  	return n;
> > > >  }
> > > >  
> > > > -static BIN_ATTR_RO(cpumap, 0);
> > > > +static BIN_ATTR_RO(cpumap, CPUMAP_FILE_MAX_BYTES);
> > > >  
> > > >  static inline ssize_t cpulist_read(struct file *file, struct kobject *kobj,
> > > >  				   struct bin_attribute *attr, char *buf,
> > > > @@ -66,7 +66,7 @@ static inline ssize_t cpulist_read(struct file *file, struct kobject *kobj,
> > > >  	return n;
> > > >  }
> > > >  
> > > > -static BIN_ATTR_RO(cpulist, 0);
> > > > +static BIN_ATTR_RO(cpulist, CPULIST_FILE_MAX_BYTES);
> > > >  
> > > >  /**
> > > >   * struct node_access_nodes - Access class device to hold user visible
> > > > diff --git a/drivers/base/topology.c b/drivers/base/topology.c
> > > > index ac6ad9ab67f9..89f98be5c5b9 100644
> > > > --- a/drivers/base/topology.c
> > > > +++ b/drivers/base/topology.c
> > > > @@ -62,47 +62,47 @@ define_id_show_func(ppin, "0x%llx");
> > > >  static DEVICE_ATTR_ADMIN_RO(ppin);
> > > >  
> > > >  define_siblings_read_func(thread_siblings, sibling_cpumask);
> > > > -static BIN_ATTR_RO(thread_siblings, 0);
> > > > -static BIN_ATTR_RO(thread_siblings_list, 0);
> > > > +static BIN_ATTR_RO(thread_siblings, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(thread_siblings_list, CPULIST_FILE_MAX_BYTES);
> > > >  
> > > >  define_siblings_read_func(core_cpus, sibling_cpumask);
> > > > -static BIN_ATTR_RO(core_cpus, 0);
> > > > -static BIN_ATTR_RO(core_cpus_list, 0);
> > > > +static BIN_ATTR_RO(core_cpus, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(core_cpus_list, CPULIST_FILE_MAX_BYTES);
> > > >  
> > > >  define_siblings_read_func(core_siblings, core_cpumask);
> > > > -static BIN_ATTR_RO(core_siblings, 0);
> > > > -static BIN_ATTR_RO(core_siblings_list, 0);
> > > > +static BIN_ATTR_RO(core_siblings, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(core_siblings_list, CPULIST_FILE_MAX_BYTES);
> > > >  
> > > >  #ifdef TOPOLOGY_CLUSTER_SYSFS
> > > >  define_siblings_read_func(cluster_cpus, cluster_cpumask);
> > > > -static BIN_ATTR_RO(cluster_cpus, 0);
> > > > -static BIN_ATTR_RO(cluster_cpus_list, 0);
> > > > +static BIN_ATTR_RO(cluster_cpus, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(cluster_cpus_list, CPULIST_FILE_MAX_BYTES);
> > > >  #endif
> > > >  
> > > >  #ifdef TOPOLOGY_DIE_SYSFS
> > > >  define_siblings_read_func(die_cpus, die_cpumask);
> > > > -static BIN_ATTR_RO(die_cpus, 0);
> > > > -static BIN_ATTR_RO(die_cpus_list, 0);
> > > > +static BIN_ATTR_RO(die_cpus, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(die_cpus_list, CPULIST_FILE_MAX_BYTES);
> > > >  #endif
> > > >  
> > > >  define_siblings_read_func(package_cpus, core_cpumask);
> > > > -static BIN_ATTR_RO(package_cpus, 0);
> > > > -static BIN_ATTR_RO(package_cpus_list, 0);
> > > > +static BIN_ATTR_RO(package_cpus, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(package_cpus_list, CPULIST_FILE_MAX_BYTES);
> > > >  
> > > >  #ifdef TOPOLOGY_BOOK_SYSFS
> > > >  define_id_show_func(book_id, "%d");
> > > >  static DEVICE_ATTR_RO(book_id);
> > > >  define_siblings_read_func(book_siblings, book_cpumask);
> > > > -static BIN_ATTR_RO(book_siblings, 0);
> > > > -static BIN_ATTR_RO(book_siblings_list, 0);
> > > > +static BIN_ATTR_RO(book_siblings, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(book_siblings_list, CPULIST_FILE_MAX_BYTES);
> > > >  #endif
> > > >  
> > > >  #ifdef TOPOLOGY_DRAWER_SYSFS
> > > >  define_id_show_func(drawer_id, "%d");
> > > >  static DEVICE_ATTR_RO(drawer_id);
> > > >  define_siblings_read_func(drawer_siblings, drawer_cpumask);
> > > > -static BIN_ATTR_RO(drawer_siblings, 0);
> > > > -static BIN_ATTR_RO(drawer_siblings_list, 0);
> > > > +static BIN_ATTR_RO(drawer_siblings, CPUMAP_FILE_MAX_BYTES);
> > > > +static BIN_ATTR_RO(drawer_siblings_list, CPULIST_FILE_MAX_BYTES);
> > > >  #endif
> > > >  
> > > >  static struct bin_attribute *bin_attrs[] = {
> > > > diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
> > > > index fe29ac7cc469..007acdb462bd 100644
> > > > --- a/include/linux/cpumask.h
> > > > +++ b/include/linux/cpumask.h
> > > > @@ -1071,4 +1071,20 @@ cpumap_print_list_to_buf(char *buf, const struct cpumask *mask,
> > > >  	[0] =  1UL							\
> > > >  } }
> > > >  
> > > > +/* 
> > > > + * Provide a valid theoretical max size for cpumap ands cpulist sysfs files to 
> > > 
> > > s/ands/and
> > > 
> > > > + * avoid breaking userspace which may allocate a buffer based on the size 
> > > > + * reported by e.g. fstat.
> > > > + *
> > > > + * For cpumap NR_CPUS/2 is a simplification of NR_CPUS/4 + NR_CPUS/32. 
> > > > + *
> > > > + * For cpulist 7 is (ceil(log10(NR_CPUS)) + 1) allowing for NR_CPUS to be up to 
> > > > + * 2 orders of magnitude larger than 8192. This covers a worst-case of every 
> > > > + * other cpu being on one of two nodes for a very large NR_CPUS.
> > > > + *
> > > > + *  Use PAGE_SIZE as a minimum for smaller configurations. 
> > > > + */
> > > > +#define CPUMAP_FILE_MAX_BYTES  (((NR_CPUS >> 1) > PAGE_SIZE) ? NR_CPUS >> 1 : PAGE_SIZE)
> > > > +#define CPULIST_FILE_MAX_BYTES  (((NR_CPUS * 7) > PAGE_SIZE) ? NR_CPUS * 7 : PAGE_SIZE)
> > > > +
> > > >  #endif /* __LINUX_CPUMASK_H */
> > > > -- 
> > > > 2.31.1
> > > 
> > 
> > -- 
> 

--