Date: Wed, 17 Nov 2010 18:35:41 +0200 From: Kostik Belousov <kostikbel@gmail.com> To: Mike Tancsa <mike@sentex.net> Cc: stable@freebsd.org Subject: Re: Call for testers: FPU changes Message-ID: <20101117163541.GR2392@deviant.kiev.zoral.com.ua> In-Reply-To: <4CE333EF.10406@sentex.net> References: <20101115211350.GE2392@deviant.kiev.zoral.com.ua> <4CE1FDBA.9030403@sentex.net> <20101116094330.GH2392@deviant.kiev.zoral.com.ua> <4CE300DE.8010304@sentex.net> <20101116221926.GN2392@deviant.kiev.zoral.com.ua> <4CE333EF.10406@sentex.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--QHdPxs9xTQFJbl9m Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Tue, Nov 16, 2010 at 08:46:23PM -0500, Mike Tancsa wrote: > On 11/16/2010 5:19 PM, Kostik Belousov wrote: > > Would your conclusion be that the patch seems to increase the throughput > > of the aesni(4) ? > >=20 > > I think that on small-sized blocks, when using aesni(4), the dominating > > factor is the copying/copyout of the data to/from the kernel address > > space. Still would be interesting to compare the full output > > of "openssl speed" on aesni(4) with and without the patch I posted. >=20 > Hi, > There does seem to be some improvement on large blocks. But there are > some freakishly fast times. On other sizes, there is no difference in > speed it would seem >=20 > I did 20 runs. Updated stats at http://www.tancsa.com/fpu.html Thank you. Indeed, I think that the test units are too small so that random system events can cause the variation. Nonetheless, patch seems to help, so I committed it. Meantime, the similar change may be beneficial for padlock(4) too. f you are going to test it, please note that most likely, openssl padlock engine does not use padlock(4), I do not know for sure. diff --git a/sys/crypto/via/padlock.c b/sys/crypto/via/padlock.c index 77e059b..ba63093 100644 --- a/sys/crypto/via/padlock.c +++ b/sys/crypto/via/padlock.c @@ -170,7 +170,7 @@ padlock_newsession(device_t dev, uint32_t *sidp, struct= cryptoini *cri) struct padlock_session *ses =3D NULL; struct cryptoini *encini, *macini; struct thread *td; - int error; + int error, saved_ctx; =20 if (sidp =3D=3D NULL || cri =3D=3D NULL) return (EINVAL); @@ -238,10 +238,18 @@ padlock_newsession(device_t dev, uint32_t *sidp, stru= ct cryptoini *cri) =20 if (macini !=3D NULL) { td =3D curthread; - error =3D fpu_kern_enter(td, &ses->ses_fpu_ctx, FPU_KERN_NORMAL); + if (!is_fpu_kern_thread(0)) { + error =3D fpu_kern_enter(td, &ses->ses_fpu_ctx, + FPU_KERN_NORMAL); + saved_ctx =3D 1; + } else { + error =3D 0; + saved_ctx =3D 0; + } if (error =3D=3D 0) { error =3D padlock_hash_setup(ses, macini); - fpu_kern_leave(td, &ses->ses_fpu_ctx); + if (saved_ctx) + fpu_kern_leave(td, &ses->ses_fpu_ctx); } if (error !=3D 0) { padlock_freesession_one(sc, ses, 0); diff --git a/sys/crypto/via/padlock_cipher.c b/sys/crypto/via/padlock_ciphe= r.c index 0ae26c8..1456ddf 100644 --- a/sys/crypto/via/padlock_cipher.c +++ b/sys/crypto/via/padlock_cipher.c @@ -205,7 +205,7 @@ padlock_cipher_process(struct padlock_session *ses, str= uct cryptodesc *enccrd, struct thread *td; u_char *buf, *abuf; uint32_t *key; - int allocated, error; + int allocated, error, saved_ctx; =20 buf =3D padlock_cipher_alloc(enccrd, crp, &allocated); if (buf =3D=3D NULL) @@ -250,14 +250,21 @@ padlock_cipher_process(struct padlock_session *ses, s= truct cryptodesc *enccrd, } =20 td =3D curthread; - error =3D fpu_kern_enter(td, &ses->ses_fpu_ctx, FPU_KERN_NORMAL); + if (!is_fpu_kern_thread(0)) { + error =3D fpu_kern_enter(td, &ses->ses_fpu_ctx, FPU_KERN_NORMAL); + saved_ctx =3D 1; + } else { + error =3D 0; + saved_ctx =3D 0; + } if (error !=3D 0) goto out; =20 padlock_cbc(abuf, abuf, enccrd->crd_len / AES_BLOCK_LEN, key, cw, ses->ses_iv); =20 - fpu_kern_leave(td, &ses->ses_fpu_ctx); + if (saved_ctx) + fpu_kern_leave(td, &ses->ses_fpu_ctx); =20 if (allocated) { crypto_copyback(crp->crp_flags, crp->crp_buf, enccrd->crd_skip, diff --git a/sys/crypto/via/padlock_hash.c b/sys/crypto/via/padlock_hash.c index 58c58b2..0fe182b 100644 --- a/sys/crypto/via/padlock_hash.c +++ b/sys/crypto/via/padlock_hash.c @@ -366,17 +366,24 @@ padlock_hash_process(struct padlock_session *ses, str= uct cryptodesc *maccrd, struct cryptop *crp) { struct thread *td; - int error; + int error, saved_ctx; =20 td =3D curthread; - error =3D fpu_kern_enter(td, &ses->ses_fpu_ctx, FPU_KERN_NORMAL); + if (!is_fpu_kern_thread(0)) { + error =3D fpu_kern_enter(td, &ses->ses_fpu_ctx, FPU_KERN_NORMAL); + saved_ctx =3D 1; + } else { + error =3D 0; + saved_ctx =3D 0; + } if (error !=3D 0) return (error); if ((maccrd->crd_flags & CRD_F_KEY_EXPLICIT) !=3D 0) padlock_hash_key_setup(ses, maccrd->crd_key, maccrd->crd_klen); =20 error =3D padlock_authcompute(ses, maccrd, crp->crp_buf, crp->crp_flags); - fpu_kern_leave(td, &ses->ses_fpu_ctx); + if (saved_ctx) + fpu_kern_leave(td, &ses->ses_fpu_ctx); return (error); } =20 --QHdPxs9xTQFJbl9m Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (FreeBSD) iEYEARECAAYFAkzkBFwACgkQC3+MBN1Mb4glLgCg84M8O4YhICTfLgjBQpS5rKsK X60AoL1rJFCn2zjHDZpihSLXdsQTU3tA =7tWt -----END PGP SIGNATURE----- --QHdPxs9xTQFJbl9m--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20101117163541.GR2392>