[clamav-users] Input Stream Scanning for very large files
clamav-users at iment.com
Fri Jan 25 13:12:25 EST 2019
I understand that it's impractical for ClamAV to scan exceedingly large
files, as it could fill up RAM and/or page forever. But the current 4GB
hard limit is overly restrictive, especially since 32-bit addresses and
numbers are ancient history in current OSes.
In particular, scanning big archives immediately after downloading is
desirable, since there can be malware that attacks the de-compressing or
de-archiving mechanism itself. Thus simply scanning the individual
contained files isn't completely adequate.
Is there any plan to allow scanning bigger files? There still are,
after all, size limits specifiable in the config file(s), and warnings
about the consequences of setting limits too big can be documented.
On Fri, 25 Jan 2019 14:32:55 +0000
"Micah Snyder (micasnyd)" <micasnyd at cisco.com> wrote:
garding specific limits, I'm sorry to say that ClamAV is presently
> limited to max file size of 4GB on most systems (and, I think
> unintentionally, 2GB on some systems).
> On Jan 24, 2019, at 4:23 PM, J.R.
> <themadbeaker at gmail.com<mailto:themadbeaker at gmail.com>> wrote:
> I think I framed my problem statement differently.
> So, our requirement is similar the one asked by John in the
> below link. I do not know if the solution proposed is a correct one..
> Also, how do you propose I should scan an archive of 100GB ( let's
> say) size. Does clamav have any limitations on scanning a single file
> of such huge size ??
> Without knowing more about this "archive" it's hard to say if ClamAV
> will even pick up anything, due to the reason Micah gave in his reply.
> But another issue is if this is just one humongous file you are trying
> to shove through and say it *does* trigger some virus... How are you
> going to know what / where the virus is? All you know is its somewhere
> in your massive archive file...
> You would be much better off scanning the individual files as you
> assemble said archive, and obviously only need to scan files where an
> infection would make sense (i.e. a text file isn't going to contain a
> There are stream settings in the clamd.conf, but I don't know what the
> hard upper-limits are.
> In cases like this, it's probably best to assemble you own sample
> archives, one clean & one infected, and run through your proposed
> process. If it works as intended, then create a few more samples and
> re-test... If it doesn't work as intended then you'll need to re-think
> your process...
More information about the clamav-users