Linux Kernel Mentees Archive on lore.kernel.org
 help / color / Atom feed
From: Amol Grover <frextrite@gmail.com>
To: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Corey Minyard <minyard@acm.org>,
	kbuild-all@lists.01.org, kbuild test robot <lkp@intel.com>,
	Arnd Bergmann <arnd@arndb.de>,
	linux-kernel@vger.kernel.org,
	Joel Fernandes <joel@joelfernandes.org>,
	openipmi-developer@lists.sourceforge.net,
	linux-kernel-mentees@lists.linuxfoundation.org
Subject: Re: [Linux-kernel-mentees] [PATCH v2] drivers: char: ipmi: ipmi_msghandler: Pass lockdep expression to RCU lists
Date: Wed, 15 Jan 2020 18:06:53 +0530
Message-ID: <20200115123653.GA20601@workstation-portable> (raw)
In-Reply-To: <20200114175828.GR2935@paulmck-ThinkPad-P72>

On Tue, Jan 14, 2020 at 09:58:28AM -0800, Paul E. McKenney wrote:
> On Tue, Jan 14, 2020 at 08:30:30AM +0530, Amol Grover wrote:
> > On Sun, Jan 12, 2020 at 01:25:58PM +0800, kbuild test robot wrote:
> > > Hi Amol,
> > > 
> > > Thank you for the patch! Perhaps something to improve:
> > > 
> > > [auto build test WARNING on char-misc/char-misc-testing]
> > > [also build test WARNING on ipmi/for-next arm-soc/for-next v5.5-rc5]
> > > [if your patch is applied to the wrong git tree, please drop us a note to help
> > > improve the system. BTW, we also suggest to use '--base' option to specify the
> > > base tree in git format-patch, please see https://stackoverflow.com/a/37406982]
> > > 
> > > url:    https://github.com/0day-ci/linux/commits/Amol-Grover/drivers-char-ipmi-ipmi_msghandler-Pass-lockdep-expression-to-RCU-lists/20200111-081002
> > > base:   https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc.git 16bb7abc4a6b9defffa294e4dc28383e62a1dbcf
> > > config: x86_64-randconfig-a003-20200109 (attached as .config)
> > > compiler: gcc-5 (Ubuntu 5.5.0-12ubuntu1) 5.5.0 20171010
> > > reproduce:
> > >         # save the attached .config to linux build tree
> > >         make ARCH=x86_64 
> > > 
> > > If you fix the issue, kindly add following tag
> > > Reported-by: kbuild test robot <lkp@intel.com>
> > > 
> > > All warnings (new ones prefixed by >>):
> > > 
> > >    In file included from include/linux/export.h:43:0,
> > >                     from include/linux/linkage.h:7,
> > >                     from include/linux/kernel.h:8,
> > >                     from include/linux/list.h:9,
> > >                     from include/linux/module.h:12,
> > >                     from drivers/char/ipmi/ipmi_msghandler.c:17:
> > >    drivers/char/ipmi/ipmi_msghandler.c: In function 'find_cmd_rcvr':
> > >    include/linux/rculist.h:53:25: warning: suggest parentheses around '&&' within '||' [-Wparentheses]
> > >      RCU_LOCKDEP_WARN(!cond && !rcu_read_lock_any_held(),  \
> > >                             ^
> > 
> > As mentioned above, RCU_LOCKDEP_WARN macro is called from
> > __list_check_rcu with 2 parameters
> > 
> > 1. !cond && !rcu_read_lock_any_held()
> > 2. The message to display incase there is a lockdep warning.
> > 
> > 
> > However, if I pass the lockdep checking condition as:
> > 
> > list_for_each_entry_rcu(ptr, list, head, lockdep_is_held(&some_lock) || rcu_read_lock_held())
> 
> Right, given the _rcu() suffix on the command, the rcu_read_lock_held()
> is implied.
> 
> > this trickles down to __list_check_rcu and then finally to
> > RCU_LOCKDEP_WARN as (here cond is `lockdep_is_held(&some_lock) || rcu_read_lock_held()`):
> > 
> > RCU_LOCKDEP_WARN(!lockdep_is_held(&some_lock) || rcu_read_lock_held() && !rcu_read_lock_any_held())
> > 
> > which according to operator precedence (I hopefully got them right)
> > would always evaluate to true if we are in an RCU read-side critical
> > section (without a lock), and hence, result in a false-positive lockdep
> > warning.
> 
> It looks that way to me.  But why not actually try it out?  After all,
> only the running system knows for sure.  And there might be some trick
> that we are both missing.
> 

I just tested this, here are the results:

Case 1: Using`lockdep_is_held() || rcu_read_lock_held()`

lock	RCU RSCS		Splat?	Actual
Y	Y			N	N
Y	N			N	N
N	Y			Y	N <=
N	N			Y	Y

Similar for
Case 2: Using `rcu_read_lock_held() || lockdep_is_held()`

Case 3: Consider 2 locks (outside rcu_read_lock())
`lockdep_is_held(lock1) || lockdep_is_held(lock2)`

lock1	lock2			Splat?	Actual
Y	Y			N*	N
Y	N			N	N
N	Y			Y	N <=
N	N			Y	Y

This too proves the hypothesis (I'd like to call that).


*However, this shows an interesting result. When both lock1 and lock2
are held, according to the hypothesis, a splat should've occured, since
the check condition (albeit faulty atm) would be:

`!lockdep_is_held(lock1) || lockdep_is_held(lock2) && !rcu_read_lock_any_held()`
=> `!T || T && !F`
=> `F || T && T`
=> `F || T`
=> `T`
However, there was no splat. Which led me to investigate further and I
found out:
1. `rcu_read_lock_any_held()` always returns 1 even if it is outside RCU
read-side CS.
2. `rcu_read_lock_held()` seems OK, returns 1 when inside and 0 when
outside

The kernel is compiled with
PROVE_RCU=y
PROVE_RCU_LIST=y

Any thoughts on this? Is this intended? And should I send-in the patch
for the first problem?

Thanks
Amol

> > This could be easily solved by putting `cond` inside brackets as it is
> > correctly done in RCU_LOCKDEP_WARN macro but not in __list_check_rcu
> > macro. Is that so, or did I miss something?
> 
> Again, that looks correct to me, but please check.
> 
> > Secondly, since there is already a condition that checks for RCU
> > read-side critical section, the extra `rcu_read_lock_held()` we supply
> > is sort of redundant and can be skipped right?
> 
> Yes, the general rule is that if the primitives ends with _rcu(), any
> lockdep condition will be in addition to rcu_read_lock_any_held().
> So you should not need to pass RCU read-side lockdep expressions to 
> primitives whose names end in _rcu..
> 
> 							Thanx, Paul
> 
> > Thanks
> > Amol
> > 
> > >    include/linux/compiler.h:58:52: note: in definition of macro '__trace_if_var'
> > >     #define __trace_if_var(cond) (__builtin_constant_p(cond) ? (cond) : __trace_if_value(cond))
> > >                                                        ^
> > > >> include/linux/rcupdate.h:263:3: note: in expansion of macro 'if'
> > >       if (debug_lockdep_rcu_enabled() && !__warned && (c)) { \
> > >       ^
> > > >> include/linux/rculist.h:53:2: note: in expansion of macro 'RCU_LOCKDEP_WARN'
> > >      RCU_LOCKDEP_WARN(!cond && !rcu_read_lock_any_held(),  \
> > >      ^
> > >    include/linux/rculist.h:371:7: note: in expansion of macro '__list_check_rcu'
> > >      for (__list_check_rcu(dummy, ## cond, 0),   \
> > >           ^
> > >    drivers/char/ipmi/ipmi_msghandler.c:1607:2: note: in expansion of macro 'list_for_each_entry_rcu'
> > >      list_for_each_entry_rcu(rcvr, &intf->cmd_rcvrs, link,
> > >      ^
> > >    include/linux/rculist.h:53:25: warning: suggest parentheses around '&&' within '||' [-Wparentheses]
> > >      RCU_LOCKDEP_WARN(!cond && !rcu_read_lock_any_held(),  \
> > >                             ^
> > >    include/linux/compiler.h:58:61: note: in definition of macro '__trace_if_var'
> > >     #define __trace_if_var(cond) (__builtin_constant_p(cond) ? (cond) : __trace_if_value(cond))
> > >                                                                 ^
> > > >> include/linux/rcupdate.h:263:3: note: in expansion of macro 'if'
> > >       if (debug_lockdep_rcu_enabled() && !__warned && (c)) { \
> > >       ^
> > > >> include/linux/rculist.h:53:2: note: in expansion of macro 'RCU_LOCKDEP_WARN'
> > >      RCU_LOCKDEP_WARN(!cond && !rcu_read_lock_any_held(),  \
> > >      ^
> > >    include/linux/rculist.h:371:7: note: in expansion of macro '__list_check_rcu'
> > >      for (__list_check_rcu(dummy, ## cond, 0),   \
> > >           ^
> > >    drivers/char/ipmi/ipmi_msghandler.c:1607:2: note: in expansion of macro 'list_for_each_entry_rcu'
> > >      list_for_each_entry_rcu(rcvr, &intf->cmd_rcvrs, link,
> > >      ^
> > >    include/linux/rculist.h:53:25: warning: suggest parentheses around '&&' within '||' [-Wparentheses]
> > >      RCU_LOCKDEP_WARN(!cond && !rcu_read_lock_any_held(),  \
> > >                             ^
> > >    include/linux/compiler.h:69:3: note: in definition of macro '__trace_if_value'
> > >      (cond) ?     \
> > >       ^
> > >    include/linux/compiler.h:56:28: note: in expansion of macro '__trace_if_var'
> > >     #define if(cond, ...) if ( __trace_if_var( !!(cond , ## __VA_ARGS__) ) )
> > >                                ^
> > > >> include/linux/rcupdate.h:263:3: note: in expansion of macro 'if'
> > >       if (debug_lockdep_rcu_enabled() && !__warned && (c)) { \
> > >       ^
> > > >> include/linux/rculist.h:53:2: note: in expansion of macro 'RCU_LOCKDEP_WARN'
> > >      RCU_LOCKDEP_WARN(!cond && !rcu_read_lock_any_held(),  \
> > >      ^
> > >    include/linux/rculist.h:371:7: note: in expansion of macro '__list_check_rcu'
> > >      for (__list_check_rcu(dummy, ## cond, 0),   \
> > >           ^
> > >    drivers/char/ipmi/ipmi_msghandler.c:1607:2: note: in expansion of macro 'list_for_each_entry_rcu'
> > >      list_for_each_entry_rcu(rcvr, &intf->cmd_rcvrs, link,
> > >      ^
> > > 
> > > vim +/if +263 include/linux/rcupdate.h
> > > 
> > > 632ee200130899 Paul E. McKenney 2010-02-22  254  
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  255  /**
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  256   * RCU_LOCKDEP_WARN - emit lockdep splat if specified condition is met
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  257   * @c: condition to check
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  258   * @s: informative message
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  259   */
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  260  #define RCU_LOCKDEP_WARN(c, s)						\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  261  	do {								\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  262  		static bool __section(.data.unlikely) __warned;		\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18 @263  		if (debug_lockdep_rcu_enabled() && !__warned && (c)) {	\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  264  			__warned = true;				\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  265  			lockdep_rcu_suspicious(__FILE__, __LINE__, s);	\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  266  		}							\
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  267  	} while (0)
> > > f78f5b90c4ffa5 Paul E. McKenney 2015-06-18  268  
> > > 
> > > :::::: The code at line 263 was first introduced by commit
> > > :::::: f78f5b90c4ffa559e400c3919a02236101f29f3f rcu: Rename rcu_lockdep_assert() to RCU_LOCKDEP_WARN()
> > > 
> > > :::::: TO: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > :::::: CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > 
> > > ---
> > > 0-DAY kernel test infrastructure                 Open Source Technology Center
> > > https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org Intel Corporation
> > 
> > 
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees

  reply index

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-10 16:47 Amol Grover
2020-01-12  5:25 ` kbuild test robot
2020-01-14  3:00   ` Amol Grover
2020-01-14 17:58     ` Paul E. McKenney
2020-01-15 12:36       ` Amol Grover [this message]
2020-01-15 19:32         ` Paul E. McKenney
2020-01-16  2:54           ` Amol Grover
2020-01-16  3:21             ` Paul E. McKenney
2020-01-16 16:27               ` Amol Grover

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200115123653.GA20601@workstation-portable \
    --to=frextrite@gmail.com \
    --cc=arnd@arndb.de \
    --cc=joel@joelfernandes.org \
    --cc=kbuild-all@lists.01.org \
    --cc=linux-kernel-mentees@lists.linuxfoundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lkp@intel.com \
    --cc=minyard@acm.org \
    --cc=openipmi-developer@lists.sourceforge.net \
    --cc=paulmck@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux Kernel Mentees Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-kernel-mentees/0 linux-kernel-mentees/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-kernel-mentees linux-kernel-mentees/ https://lore.kernel.org/linux-kernel-mentees \
		linux-kernel-mentees@lists.linuxfoundation.org
	public-inbox-index linux-kernel-mentees

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.linuxfoundation.lists.linux-kernel-mentees


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git