Security Vulnerability: Spectre side channel attack "Bounds Check Bypass Store" aka CVE-2018-3693.

  • 7023075
  • 11-Jun-2018
  • 15-Feb-2019


SUSE Linux Enterprise Server 12 Service Pack 3 (SLES 12 SP3)
SUSE Linux Enterprise Server 11 Service Pack 4 (SLES 11 SP4)


An improved Spectre side channel style attack has been identified, called "Bounds Check Bypass Store".
This issue is similar in behavior to Spectre Variant 1 and Variant 2.

Code patterns storing an attacker controlled value (that would ultimately fail due to a bounds check, or lack of write permissions), could leak this value during speculative execution as reads that will already "see" the value as being stored, and so could generate side-channel information leaks.

For instance, memory stores that have overflow checks before them, are still speculatively executed with stores happening over the ends of the buffer into other variables, that in turn could be used speculatively.

In theory, Spectre v1 style attack values could persist even longer in the processors speculation queue when stored to memory temporary.

Additionally, a possible approach for this attack, is that speculative buffer overwrites of for example return address pointers,  could lead to mispredicted returns, and so further attacker controlled speculatively executed code patterns, similar to Spectre Variant 2 may be executed.

Please note :
All happens in the speculative execution phase, and nothing will be written to memory in the end.
Only information leaks via side-channels are possible with these type of attacks.


The software mitigations that are needed for this issue are similar to the ones for Spectre Variant 1, and in fact, the Spectre v1 mitigations SUSE has applied cover this issue to a large part already.

Please see Spectre Variant 1 details here : CVE-2017-5753

Further speculative fencing will be added in affected places throughout the Linux Kernel and hypervisor code.


Additional Information

Research is currently undergoing by SUSE and upstream developers to locate and add fencing to affected places of code.

Note :
- These mitigations do not need CPU microcode fixes.
- These mitigations have only very limited performance impact.
   Potential performance loss largely depends on systemcall load / patterns.

Feedback service temporarily unavailable. For content questions or problems, please contact Support.