Quadratic runtime with malformed PDF missing xref marker

[Google-Autofuzz]

When running the following code with PyPDF2==1.27.8 on the attached input results in 100% CPU consumption for an arbitrary long time.

MCVE

PDF: test.pdf

from PyPDF2 import PdfFileReader

reader = PdfFileReader("test.pdf")

[MartinThoma]

The PDF is hopelessly broken:

$ cat issue-582.pdf 
ang��H
%s/I\ʱ\\\\\\\\\\\\\\\\\\\\\\\\\[\\𝟖\\\
%EO-254536430
%%EOFF�	g
e�
%%EOF%                         

Investigating

import time
import PyPDF2
import traceback

print(PyPDF2.__version__)

t0 = time.time()
try:
    reader = PyPDF2.PdfFileReader("test.pdf")
except Exception as e:
    print(traceback.format_exc())
    print(e)
t1 = time.time()
print(f"`PyPDF2=={PyPDF2.__version__}` finished after {t1-t0:.3f}s")

Affected Versions

• PyPDF2==1.20 finished after 24.664s
• PyPDF2==1.25 finished after 29.781s
• PyPDF2==1.26.0 finished after 29.743s
• PyPDF2==1.27.0 finished after 29.611s
• PyPDF2==1.27.8 finished after 34.518s
• Fixed in c6c56f5 via fix pdf reader getting stuck when trying to read large files wihhout xref marker #808
• PyPDF2==1.27.9: Exception after 0.001

[MartinThoma]

Reported publicly via GHSA-jrm6-h9cq-8gqw