Re: [bug] crypto/vmx/p8_ghash memory corruption in 4.8-rc7
flat view
From: Jan Stancek <jstancek@redhat.com>
Date: 2016-09-28 08:59:54
Also in:
linux-crypto, lkml
Jan, Can you check if the problem occurs with this patch?
No issues in over-night test with this patch.
quoted hunk ↗ jump to hunk
--- a/drivers/crypto/vmx/vmx.c +++ b/drivers/crypto/vmx/vmx.c@@ -28,6 +28,8 @@ #include <asm/cputable.h> #include <crypto/internal/hash.h> +int p8_ghash_fallback_descsize(void); + extern struct shash_alg p8_ghash_alg; extern struct crypto_alg p8_aes_alg; extern struct crypto_alg p8_aes_cbc_alg;@@ -45,6 +47,7 @@ int __init p8_init(void) { int ret = 0; struct crypto_alg **alg_it; + int ghash_descsize; for (alg_it = algs; *alg_it; alg_it++) { ret = crypto_register_alg(*alg_it);@@ -59,6 +62,12 @@ int __init p8_init(void) if (ret) return ret; + ghash_descsize = p8_ghash_fallback_descsize(); + if (ghash_descsize < 0) { + printk(KERN_ERR "Cannot get descsize for p8_ghash fallback\n"); + return ghash_descsize; + } + p8_ghash_alg.descsize += ghash_descsize; ret = crypto_register_shash(&p8_ghash_alg); if (ret) { for (alg_it = algs; *alg_it; alg_it++)
I'd suggest to move this inside vmx/ghash.c to a new function, so all p8_ghash related code is at single place. Then p8_init() would just call the new function: - ret = crypto_register_shash(&p8_ghash_alg); + ret = register_p8_ghash(); Regards, Jan