Re: [PATCH v3 03/23] mfd: kempld: Simplify device abstraction
Uwe Kleine-König (The Capable Hub) <[email protected]>
| Newsgroups | dev.linux.lists.mfd,dev.linux.lists.sashiko-reviews |
|---|---|
| Message-ID | <alAFPoJkhMKP13FO@monoceros> |
Hello, On Thu, Jul 09, 2026 at 05:07:03PM +0000, [email protected] wrote: > > - ret = platform_device_add_data(pdev, pdata, sizeof(*pdata)); > > + ret = platform_device_add_data(pdev, pdata, sizeof(pdata)); > > [Severity: High] > Does this code pass the size of the pointer instead of the size of the > struct? > > Looking at kempld_probe(), pdata is a pointer to the static > kempld_platform_data_generic structure. After the changes in this patch, the > struct only contains a single int. > > Since sizeof(pdata) evaluates to the pointer size (typically 8 bytes), could > this cause kmemdup() inside platform_device_add_data() to read past the end > of the structure and trigger a KASAN global-out-of-bounds read during ACPI > platform device probe? This is a real issue. Will address in v4. But I wait a bit before sending that. Best regards Uwe
signature.asc
(application/pgp-signature, 488 B)
-----BEGIN PGP SIGNATURE----- iQEzBAABCgAdFiEEP4GsaTp6HlmJrf7Tj4D7WH0S/k4FAmpQBXgACgkQj4D7WH0S /k4RPwgAmYVSgTTgDqfIit7NF7lQ6cM1mSeCrDBWa9q+UuB4LP3DQ5/zoTuBZjk5 S06QoOT4Agdw49etj5uFfmfYWq4zKB7NNygAkQqTwyxFUYPIzWjtOd0apVK3ijLL 2Z+p/aPPM1o0vhZFp6t106aHJlhhzrkFI88X0GDEWBnQgL2TVgzdxKiRzeXlqNxU +iJVJBZb2IFMvwtj59wxIB7+q0RZwDJN7yTR/sGGn8Rv5bEZYbwUT2ecBouJJKNU E1R0RJfw3N2SjPW5Rs7sYfgrgio/3Y41yimbNgHyz1rWVuyhmWHo8d9i6Wdma3Dk ljXknnLpm1glQOjEnqc3n8F3bIsXIg== =FjBW -----END PGP SIGNATURE-----