[PATCH] cswift engine code and signals (e.g. SIGALRM in speed.c)

Hi all, While performing performance tests on a Sun Crypto Accelerator I Board (also known as CryptoSwift) on sparc-sun-solaris architecture, I noticed a race condition between the libswift.so-hardware-I/O and the alarm signal speed.c uses. The symptom was the openssl speed program terminating

Re: [PATCH] cswift engine code and signals (e.g. SIGALRM in speed.c)

I didn't see which version of Solaris you were running, but be aware that in 2.8, prior to KJP 12, there are some known kernel and libaio issues that can result in the signal mask being changed out from under a process. Eric Laroche wrote: Hi all, While performing performance tests on a

Re: [PATCH] cswift engine code and signals (e.g. SIGALRM in speed.c)

Have you reported this to Sun? Cheers, Ben. Eric Laroche wrote: Hi all, While performing performance tests on a Sun Crypto Accelerator I Board (also known as CryptoSwift) on sparc-sun-solaris architecture, I noticed a race condition between the libswift.so-hardware-I/O and the alarm

Re: [PATCH] cswift engine code and signals (e.g. SIGALRM in speed.c)

Hi Ben, Ben Laurie wrote: Have you reported this to Sun? Yes, I reported it to our local Sun contact person. Maybe someone has Sun's most efficient bug-fix-report e-mail address at hand? [Even the software package's contact info states: 'Please contact your local service provider'...]