Suricata process dying due to hyperscan problem

bmeeks

The Pattern Matcher Algorithm is not critical to whether detection works or not. It only matters in terms of speed versus memory consumption.

All fhe pattern matcher does is execute a regular expression (regex) search against packet data looking for whatever pattern the rule specified. The different algorithms all accomplish the exact same thing. It's just that some use less memory, but may run slower. Others may be lightning fast, but that speed comes from very high RAM consumption.

Hyperscan is a specialized regex library created by Intel to take advantage of certain complex math instruction op-codes present in their CPUs. Those specialized instructions that most Intel CPUs have allow for highly optimized regex searches. But because Hyperscan depends on instruction op-codes only present in Intel CPUs, the library does not work with ARM or any other non-Intel hardware.

The Auto setting for the Pattern Matcher algorithm in Suricata will default to the Hyperscan library if it is present. Otherwise it will choose AC (Aho-Corasick). When the Suricata binary is compiled on the Netgate package builder, it automatically senses which type of CPU is targeted (Intel or something else), and then includes the Hyperscan library only for Intel CPU targets. This means if you have a Netgate firewall appliance with an ARM chip, the Hyperscan library is not even present on your system.

gfeiner

FYI. 23.09.1 released with this memory allocation fix: https://www.freebsd.org/security/advisories/FreeBSD-EN-23:20.vm.asc

paulp

@gfeiner Unfortunately, in my case at least, Suricata still dies after this upgrade.

kiokoman

@paulp said in Suricata process dying due to hyperscan problem:

@gfeiner Unfortunately, in my case at least, Suricata still dies after this upgrade.

me too, still crashing

tgm-its

Hi all.
Just a question? Will the new update Version 2.7.2 of pFsense CE fix the Hyperscan error in Suricata?
If so I could update my Firewall if not I'll wait.

BR
Paal B.

tylerevers

Not sure what to make of this, but within 24 hours of updating to 23.09.1 one of my Interfaces in Suricata disappeared by itself. For reference, it was the only Interface I was having the Hyperscan issue with and it was running on AC-KS for weeks without an issue. It is the only one of three Interfaces that was using AC-KS instead of Auto.

I'm not quite sure how this happened. I woke up today and was just looking at systems to discover that one specific Suricata Interface had deleted itself.

bmeeks

@tylerevers said in Suricata process dying due to hyperscan problem:

I'm not quite sure how this happened. I woke up today and was just looking at systems to discover that one specific Suricata Interface had deleted itself.

There is no mechanism within the package code for that to happen autonomously.

I would check first the pfSense system log on the impacted system to see what may be logged there, then look in the configuration backup history of pfSense to see if someone took an action. The GUI code logs an appropriate message for all configuration changes saved to the config.xml fiile. You can find the configuration entire history under DIAGNOSTICS > BACKUP AND RESTORE.

bmeeks

Fingers crossed, but I think I found the bug.

I definitely found an errant double free() of memory when processing IPv4 addresses in a Pass List. When you execute a double free() of memory you will get random crashes.

Since I have been unable to reproduce the problem, I can't say for sure what I found will fix the Hyperscan issue, but I am hopeful based on the fact several of you have stated that turning off Legacy Blocking Mode (in other words, running in plain IDS mode) allows Suricata to run with no issue. The double free() was in the custom Legacy Blocking Module, and it was located in new code that was added with the first 7.0.0 Suricata update back when 23.09 Plus was still in development mode and 2.8 CE snapshots were active. Disabling Legacy Blocking Mode means this buggy portion of the module's code is not executed.

I will create a pull request and get this fix posted for Netgate to review and merge. That will not happen until the first of the coming week.

Update: the fix is posted at https://github.com/pfsense/FreeBSD-ports/pull/1333 for review and merge by the Netgate developer team.

kiokoman

@bmeeks
just out of curiosity, what's the difference between "free" and "SCfree" ? i understand that free() it's to deallocates the memory but i can't find reference for scfree

bmeeks

@kiokoman said in Suricata process dying due to hyperscan problem:

@bmeeks
just out of curiosity, what's the difference between "free" and "SCfree" ?

Not a thing currently. The upstream Suricata developers just wrap some common C functions with their own names in case they might ever want to customize them for some reason. Today the two are exactly the same. Here are the #define preprocessor definitions currently in use by upstream:

#define SCMalloc malloc
#define SCCalloc calloc
#define SCRealloc realloc
#define SCFree free

I just fixed up my code to stay in sync. It was an overlooked typo thing that I noticed while scrutinizing the code for any possible bug.

Line 718 in the GitHub link I posted is where the errant double free() call happened. Notice the new revision deletes that line.

Line 676 was a misplaced continue statement that could result in a memory leak because it bypasses the free() call to dump the IPv4 address structure created and passed to us elsewhere in the code. Notice the continue statement was moved to be after the free() call.

The other changes are just cosmetic.

kiokoman

@bmeeks
now i understand, thanks for the explanation

tylerevers

@bmeeks said in Suricata process dying due to hyperscan problem:

@tylerevers said in Suricata process dying due to hyperscan problem:

I'm not quite sure how this happened. I woke up today and was just looking at systems to discover that one specific Suricata Interface had deleted itself.

There is no mechanism within the package code for that to happen autonomously.

I would check first the pfSense system log on the impacted system to see what may be logged there, then look in the configuration backup history of pfSense to see if someone took an action. The GUI code logs an appropriate message for all configuration changes saved to the config.xml fiile. You can find the configuration entire history under DIAGNOSTICS > BACKUP AND RESTORE.

Thank you for your guidance. The configuration history indicated that my specific user made the change. My apologies for the red herring.

bmeeks

My pull request containing the anticipated fix for this Hyperscan error has been merged. An updated Suricata package has built and should appear as an available update for 2.7.2 CE and 23.09.1 Plus users.

Look for an update to version 7.0.2_2 for the Suricata package. When installed, the new package should pull in version 7.0.2_5 of the Suricata binary.

Fingers crossed this fixes the Hyperscan issue. But as I mentioned previously, since I could never reproduce the error in my small test environment, I can't say with 100% certainty the bug I found and fixed is the actual Hyperscan culprit.

NRgia

@bmeeks said in Suricata process dying due to hyperscan problem:

My pull request containing the anticipated fix for this Hyperscan error has been merged. An updated Suricata package has built and should appear as an available update for 2.7.2 CE and 23.09.1 Plus users.

Look for an update to version 7.0.2_2 for the Suricata package. When installed, the new package should pull in version 7.0.2_5 of the Suricata binary.

For 23.09.1 I can confirm that it is available.
After the update I can see these packages

pfSense-pkg-suricata-7.0.2_2   pfSense package suricata
suricata-7.0.2_5               High Performance Network IDS, IPS and Security Monitoring engine

Thank you

kiokoman

@bmeeks
tested

and .....

not working..

[340341 - RX#01-vmx2] 2023-12-11 22:42:50 Info: checksum: No packets with invalid checksum, assuming checksum offloading is NOT used
[340346 - W#05] 2023-12-11 22:42:53 Error: spm-hs: Hyperscan returned fatal error -1.
[340347 - W#06] 2023-12-11 22:42:53 Error: spm-hs: Hyperscan returned fatal error -1.

bmeeks

@kiokoman said in Suricata process dying due to hyperscan problem:

@bmeeks
tested

and .....

not working..

[340341 - RX#01-vmx2] 2023-12-11 22:42:50 Info: checksum: No packets with invalid checksum, assuming checksum offloading is NOT used
[340346 - W#05] 2023-12-11 22:42:53 Error: spm-hs: Hyperscan returned fatal error -1.
[340347 - W#06] 2023-12-11 22:42:53 Error: spm-hs: Hyperscan returned fatal error -1.

Well, crap! I had high hopes.

Does it still work if you disable blocking mode?

kiokoman

@bmeeks
yes, it's running on an interface i have without blocking mode

bmeeks

@kiokoman said in Suricata process dying due to hyperscan problem:

@bmeeks
yes, it's running on an interface i have without blocking mode

Please share the output of this command run from a shell prompt:

pkg info | grep suricata

Let's make sure you have the latest binary. It should show suricata-7.0.2_5.

kiokoman

@bmeeks

Shell Output - pkg info | grep suricata
pfSense-pkg-suricata-7.0.2_2 pfSense package suricata
suricata-7.0.2_5 High Performance Network IDS, IPS and Security Monitoring engine

this is pfsense 2.7.2
the strange part is that i have another pfsense 23.09.1-RELEASE that it's running on vmware with the same packages and more vlans but it have no trouble ....
the only difference is that 23.09.1 is running with 4 cpu and 2.7.2 with 8 cpu
tomorrow i can try to lower the cpu and see if there is any difference

Negan

Still broken here also..... :-(

[177766 - RX#01-ix0] 2023-12-12 06:48:02 Info: checksum: More than 1/10th of packets have an invalid checksum, assuming checksum offloading is used (193/1000)
[177768 - W#02] 2023-12-12 07:17:29 Error: spm-hs: Hyperscan returned fatal error -1.