Added a static Boolean on the SocketOptions class to avoid using the …

[rickshaw5724]

…opcode injection technique for reading the time stamp counter.

Fixes #1071

[codecov]

Codecov Report

Merging #1072 (26b0253) into master (f416a78) will increase coverage by 65.82%.
The diff coverage is 0.00%.

❗ Current head 26b0253 differs from pull request most recent head 1a2314a. Consider uploading reports for the commit 1a2314a to get more accurate results

@@             Coverage Diff             @@
##           master    #1072       +/-   ##
===========================================
+ Coverage        0   65.82%   +65.82%     
===========================================
  Files           0      146      +146     
  Lines           0     9074     +9074     
  Branches        0     1450     +1450     
===========================================
+ Hits            0     5973     +5973     
- Misses          0     2500     +2500     
- Partials        0      601      +601     

Files Changed	Coverage
src/NetMQ/Core/Utils/OpCode.cs	0.00%
src/NetMQ/SocketOptions.cs	ø

[drewnoakes]

This doesn't really seem like a socket option, and we don't have any other static options here as far as I can tell.

Also the timing of setting this value is important, as the result of Open is cached in Clock.s_rdtscSupported.

Because of the timing issue, what about using something like an environment variable to control this. For example, if NETQM_SUPPRESS_RDTSC was present, it would disable any attempt to call Opcode.Open() and store a false in s_rdtscSupported. All that could happen in the static constructor of Clock.

[rickshaw5724]

I agree SocketOptions isn't the right place. I'll change it to look for an environment variable

…

On Tue, Sep 5, 2023, 9:44 PM Drew Noakes ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In src/NetMQ/SocketOptions.cs <#1072 (comment)>: > + /// <summary> + /// If set, the time stamp counter is not read directly through opcode injection, + /// rather <see cref="System.Diagnostics.Stopwatch.GetTimestamp"/> is used. + /// When false, the time stamp counter is read by allocating a few bytes on the heap with + /// read/write/execute privilege. OpCode is copied to this allocated memory and invoked to read + /// the time stamp counter, (which is a register available on most modern CPUs). While this is + /// an accurate way to read the time stamp counter, because it injects code onto the heap, this + /// can be detected as a malware technique by some anti-virus defenders. + /// </summary> + public static bool DoNotUseRDTSC; This doesn't really seem like a socket option, and we don't have any other static options here as far as I can tell. Also the timing of setting this value is important, as the result of Open is cached in Clock.s_rdtscSupported. Because of the timing issue, what about using something like an environment variable to control this. For example, if NETQM_SUPPRESS_RDTSC was present, it would disable any attempt to call Opcode.Open() and store a false in s_rdtscSupported. All that could happen in the static constructor of Clock. — Reply to this email directly, view it on GitHub <#1072 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ATJKVZLWRNKPIX3DGIOOBZDXY7IP3ANCNFSM6AAAAAA4F76BBU> . You are receiving this because you authored the thread.Message ID: ***@***.***>