RFR(S): 8050972: Concurrency problem in PcDesc cache
martin.doerr at sap.com
Fri Jul 18 08:34:46 UTC 2014
yes, volatile is enough. There’s no requirement on the ordering.
The requirement for the writers is that they must only write valid entries or NULL.
The requirement for the readers is that they must not read several times for the matching and the returning of the result.
If a reader doesn’t find a value in the cache, the slow path will be used.
From: Vitaly Davidovich [mailto:vitalyd at gmail.com]
Sent: Donnerstag, 17. Juli 2014 18:39
To: Doerr, Martin
Cc: hotspot-dev developers; Vladimir Kozlov; Lindenmaier, Goetz
Subject: RE: RFR(S): 8050972: Concurrency problem in PcDesc cache
Is volatile enough though if the entries are read/written concurrently? What about needing, e.g., store-store barriers when writing an entry into the array?
Sent from my phone
On Jul 17, 2014 11:20 AM, "Doerr, Martin" <martin.doerr at sap.com<mailto:martin.doerr at sap.com>> wrote:
the following line should also work:
PcDesc* volatile _pc_descs[cache_size];
But we thought that the typedef would improve readability.
The array elements must be volatile, not the PcDescs which are pointed to.
From: hotspot-dev [mailto:hotspot-dev-bounces at openjdk.java.net<mailto:hotspot-dev-bounces at openjdk.java.net>] On Behalf Of Vladimir Kozlov
Sent: Donnerstag, 17. Juli 2014 17:09
To: Lindenmaier, Goetz; hotspot-dev at openjdk.java.net<mailto:hotspot-dev at openjdk.java.net>
Subject: Re: RFR(S): 8050972: Concurrency problem in PcDesc cache
What is the reason for new typedef?
On 7/17/14 1:54 AM, Lindenmaier, Goetz wrote:
> This webrev fixes an important concurrency issue in nmethod.
> Please review and test this change. I please need a sponsor.
> This should be fixed into 8u20, too.
> The entries of the PcDesc cache in nmethods are not declared as volatile, but they are accessed and modified by several threads concurrently. Some compilers (namely xlC 12 on AIX) duplicate some memory accesses to non-volatile fields. In this case, this has led to the situation that a thread had successfully matched a pc in the cache, but returned the reloaded value which was already overwritten by another thread.
> Best regards,
> Martin and Goetz.
More information about the hotspot-dev