CVE-2026-97589 PUBLISHED

s390/crypto: Fix wrong return code to engine in asynch callbacks

Assigner: Linux
Reserved: 24.09.2026 Published: 25.09.2026 Updated: 25.09.2026

In the Linux kernel, the following vulnerability has been resolved:

s390/crypto: Fix wrong return code to engine in asynch callbacks

When crypto_finalize_hash_request() or crypto_finalize_skcipher_request() explicitly completes a request, the do_one_request callback must return 0 to indicate successful handling. Returning a negative error code causes the crypto engine to assume the driver failed to take ownership and triggers a second completion via crypto_request_complete(), resulting in a double completion. This pattern occurs in paes_s390.c 4 times and once in phmac_s390.c.

Fixed in phmac_do_one_request() and all four paes do_one_request callbacks (ecb, cbc, ctr, xts) by returning 0 after explicit finalization instead of propagating the error code.

Metrics

CVSS Vector: CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 7

AV:L - The faulty return is in ecb/cbc/ctr/xts_paes_do_one_request() and phmac_do_one_request(), which run in the crypto engine worker. They are reached through local crypto API users: AF_ALG sockets binding "ecb(paes)" etc. (algif_skcipher recvmsg -> ecb_paes_crypt -> engine) or dm-crypt. No remote peer supplies the data that causes the failure. AC:H - The double completion fires only when the engine path returns an error. That needs the protected key to be invalidated (-EKEYEXPIRED, e.g. the wrapping key changed) and then re-conversion to fail in paes_convert_key() because the card is busy or offline. AF_ALG blocks setkey while op sockets exist, so the attacker cannot force this state. PR:L - AF_ALG socket creation and binding the paes/phmac algorithms need no capability, so a local unprivileged user holding usable key material can push requests through the paes engine path. UI:N - No victim action is needed. Once the engine-side failure condition exists, the attacker's own AF_ALG requests (or ordinary dm-crypt I/O) hit the double completion. S:U - The corruption stays within the kernel's own security authority. There is no guest-to-host or other boundary crossing. C:H - crypto_pump_requests() calls crypto_request_complete() a second time after crypto_finalize_skcipher_request(). In AF_ALG the extra complete of ctx->wait lets a later request's areq be freed while the engine still processes it, so the engine keeps reading freed memory (use-after-free). I:H - The engine's skcipher walk and cpacf output keep writing into the freed request and scatterlist memory. For dm-crypt, the same bio completes twice. Both are use-after-free writes that can be used for heap corruption. A:H - The double completion corrupts request/bio lifetime (double bio end or a use-after-free on the areq), which can crash the kernel or corrupt dm-crypt I/O.

Product Status

Vendor Linux
Product Linux
Versions Default: unaffected
  • affected from 6cd87cb5ef6ca50ae17c371482ceaab1d635e232 to 972e0d9b7d1d112240ccde1f1be85bf9ffaa1720 (excl.)
  • affected from 6cd87cb5ef6ca50ae17c371482ceaab1d635e232 to 5480291aa848e19e61175abfd457933516db70c6 (excl.)
  • affected from 6cd87cb5ef6ca50ae17c371482ceaab1d635e232 to ac1481320110b803ab9b79ab4d2ca11a74fc05f2 (excl.)
Vendor Linux
Product Linux
Versions Default: affected
  • Version 6.16 is affected
  • unaffected from 0 to 6.16 (excl.)
  • unaffected from 6.18.53 to 6.18.* (incl.)
  • unaffected from 7.2.7 to 7.2.* (incl.)
  • unaffected from 7.3-rc3 to * (incl.)

References