On 4/5/2020 2:50 PM, Iuliana Prodan wrote: > Here's the KASAN report: > BUG: KASAN: use-after-free in aead_crypt_done+0x60/0xd8 > Read of size 1 at addr ffff00002303f014 by task swapper/0/0 > > CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.6.0-rc1-00163-gd88dd5c-dirty #18 This is no a public SHA1, you are probably running with patch 1/4 applied in the tree (and with a "dirty tree"). > Hardware name: LS1046A RDB Board (DT) > Call trace: > dump_backtrace+0x0/0x260 > show_stack+0x14/0x20 > dump_stack+0xe8/0x144 > print_address_description.isra.11+0x64/0x348 > __kasan_report+0x11c/0x230 > kasan_report+0xc/0x18 > __asan_load1+0x5c/0x68 > aead_crypt_done+0x60/0xd8 > caam_jr_dequeue+0x390/0x608 > ... You should provide full KASan log - shawdow bits etc. > @@ -973,8 +973,6 @@ static void aead_crypt_done(struct device *jrdev, u32 *desc, u32 err, > > aead_unmap(jrdev, edesc, req); > > - kfree(edesc); > - > /* > * If no backlog flag, the completion of the request is done > * by CAAM, not crypto engine. > @@ -983,6 +981,8 @@ static void aead_crypt_done(struct device *jrdev, u32 *desc, u32 err, > aead_request_complete(req, ecode); > else > crypto_finalize_aead_request(jrp->engine, req, ecode); > + > + kfree(edesc); I think it's better freeing all resources before calling the completion callback, to avoid unnecessary memory strains. Horia