Impact
Passing either ‘infinity’, ‘inf’ or float(‘inf’) (or their negatives) to datetime or date fields causes validation to run forever with 100% CPU usage (on one CPU).
Patches
Pydantic is be patched with fixes available in the following versions:
v1.8.2
v1.7.4
v1.6.2
All these versions are available on pypi, and will be available on conda-forge soon.
See the changelog for details.
Workarounds
If you absolutely can’t upgrade, you can work around this risk using a validator to catch these values, brief demo:
from datetime import date
from pydantic import BaseModel, validator
class DemoModel(BaseModel):
date_of_birth: date
@validator('date_of_birth', pre=True)
def skip_infinite_values(cls, v):
try:
seconds = float(v)
except (ValueError, TypeError):
return v
else:
if seconds == float('inf'):
return date.max
elif seconds == float('-inf'):
return date.min
else:
return seconds
Note: this is not an ideal solution (in particular you’ll need a slightly different function for datetimes), instead of a hack like this you should upgrade pydantic.
If you are not using v1.8.x, v1.7.x or v1.6.x and are unable to upgrade to a fixed version of pydantic, please create an issue requesting a back-port, and we will endeavour to release a patch for earlier versions of pydantic.
References
This was fixed in commit 7e83fdd.
github.com/pydantic/pydantic/commit/1c24f1d74ba95ea985b50bdc001ce96c813229aa
github.com/pydantic/pydantic/commit/80e0dd3f752bef145dce12f160d262bb40ec8d47
github.com/pydantic/pydantic/commit/bdde15b7b947c94ca00fd6eb92da8db390a13520
github.com/samuelcolvin/pydantic/commit/7e83fdd2563ffac081db7ecdf1affa65ef38c468
github.com/samuelcolvin/pydantic/security/advisories/GHSA-5jqp-qgf6-3pvh
lists.fedoraproject.org/archives/list/[email protected]/message/S2HT266L6Q7H6ICP7DFGXOGBJHNNKMKB
lists.fedoraproject.org/archives/list/[email protected]/message/UEFWM7DYKD2ZHE7R5YT5EQWJPV4ZKYRB
lists.fedoraproject.org/archives/list/[email protected]/message/UMKAJX4O6IGBBCE32CO2G7PZQCCQSBLV
nvd.nist.gov/vuln/detail/CVE-2021-29510