Got tons of great test results but still not an independent RAID 1
rebuild. I really want to commit this but I won't until someone else
besides me tests this.
On Thu, Oct 21, 2010 at 02:17:58PM -0500, Marco Peereboom wrote:
> anyone?
>
> On Wed, Oct 20, 2010 at 08:47:00PM -0500, Marco Peereboom wrote:
> > On Thu, Sep 30, 2010 at 03:35:33AM +0200, Tobias Ulmer wrote:
> > > I got this after a while:
> > >
> > > panic: softraid0: sr_crypto_finish_io
> > >
> > > No serial, so there's no more info. You know where to find me
> >
> > new diff that should fix all them issues.
> >
> > please test, especially raid 1 including rebuild and stuff.
> >
> > Index: dev/softraid.c
> > ===================================================================
> > RCS file: /cvs/src/sys/dev/softraid.c,v
> > retrieving revision 1.215
> > diff -u -p -r1.215 softraid.c
> > --- dev/softraid.c 12 Oct 2010 00:53:32 -0000 1.215
> > +++ dev/softraid.c 21 Oct 2010 01:36:32 -0000
> > @@ -126,6 +126,7 @@ void sr_rebuild(void *);
> > void sr_rebuild_thread(void *);
> > void sr_roam_chunks(struct sr_discipline *);
> > int sr_chunk_in_use(struct sr_softc *, dev_t);
> > +void sr_startwu_callback(void *, void *);
> >
> > /* don't include these on RAMDISK */
> > #ifndef SMALL_KERNEL
> > @@ -1806,6 +1807,8 @@ sr_wu_put(struct sr_workunit *wu)
> > wu->swu_fake = 0;
> > wu->swu_flags = 0;
> >
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_wu_put", DEVNAME(sd->sd_sc));
> > while ((ccb = TAILQ_FIRST(&wu->swu_ccb)) != NULL) {
> > TAILQ_REMOVE(&wu->swu_ccb, ccb, ccb_link);
> > sr_ccb_put(ccb);
> > @@ -2563,6 +2566,9 @@ sr_hotspare_rebuild(struct sr_discipline
> > busy = 0;
> >
> > s = splbio();
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_hotspare_rebuild",
> > + DEVNAME(sd->sd_sc));
> > TAILQ_FOREACH(wu, &sd->sd_wu_pendq, swu_link) {
> > TAILQ_FOREACH(ccb, &wu->swu_ccb, ccb_link) {
> > if (ccb->ccb_target == chunk_no)
> > @@ -2816,6 +2822,11 @@ sr_ioctl_createraid(struct sr_softc *sc,
> > sd = malloc(sizeof(struct sr_discipline), M_DEVBUF, M_WAITOK | M_ZERO);
> > sd->sd_sc = sc;
> > SLIST_INIT(&sd->sd_meta_opt);
> > + sd->sd_workq = workq_create("srdis", 1, IPL_BIO);
> > + if (sd->sd_workq == NULL) {
> > + printf("%s: could not create workq\n");
> > + goto unwind;
> > + }
> > if (sr_discipline_init(sd, bc->bc_level)) {
> > printf("%s: could not initialize discipline\n", DEVNAME(sc));
> > goto unwind;
> > @@ -3407,6 +3418,9 @@ sr_discipline_shutdown(struct sr_discipl
> >
> > sr_chunks_unwind(sc, &sd->sd_vol.sv_chunk_list);
> >
> > + if (sd->sd_workq)
> > + workq_destroy(sd->sd_workq);
> > +
> > if (sd)
> > sr_discipline_free(sd);
> >
> > @@ -3625,10 +3639,29 @@ sr_raid_sync(struct sr_workunit *wu)
> > }
> >
> > void
> > +sr_startwu_callback(void *arg1, void *arg2)
> > +{
> > + struct sr_discipline *sd = arg1;
> > + struct sr_workunit *wu = arg2;
> > + struct sr_ccb *ccb;
> > + int s;
> > +
> > + s = splbio();
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_startwu_callback", DEVNAME(sd->sd_sc));
> > + wu->swu_cb_active = 1;
> > +
> > + TAILQ_FOREACH(ccb, &wu->swu_ccb, ccb_link)
> > + VOP_STRATEGY(&ccb->ccb_buf);
> > +
> > + wu->swu_cb_active = 0;
> > + splx(s);
> > +}
> > +
> > +void
> > sr_raid_startwu(struct sr_workunit *wu)
> > {
> > struct sr_discipline *sd = wu->swu_dis;
> > - struct sr_ccb *ccb;
> >
> > splassert(IPL_BIO);
> >
> > @@ -3643,9 +3676,8 @@ sr_raid_startwu(struct sr_workunit *wu)
> > TAILQ_INSERT_TAIL(&sd->sd_wu_pendq, wu, swu_link);
> >
> > /* start all individual ios */
> > - TAILQ_FOREACH(ccb, &wu->swu_ccb, ccb_link) {
> > - VOP_STRATEGY(&ccb->ccb_buf);
> > - }
> > + workq_queue_task(sd->sd_workq, &wu->swu_wqt, 0, sr_startwu_callback,
> > + sd, wu);
> > }
> >
> > void
> > Index: dev/softraid_crypto.c
> > ===================================================================
> > RCS file: /cvs/src/sys/dev/softraid_crypto.c,v
> > retrieving revision 1.57
> > diff -u -p -r1.57 softraid_crypto.c
> > --- dev/softraid_crypto.c 27 Sep 2010 19:49:43 -0000 1.57
> > +++ dev/softraid_crypto.c 5 Oct 2010 20:49:24 -0000
> > @@ -164,11 +164,11 @@ sr_crypto_create(struct sr_discipline *s
> >
> > } else if (sr_crypto_get_kdf(bc, sd))
> > goto done;
> > -
> > +
> > /* Passphrase volumes cannot be automatically assembled. */
> > if (!(bc->bc_flags & BIOC_SCNOAUTOASSEMBLE) && bc->bc_key_disk == NODEV)
> > goto done;
> > -
> > +
> > strlcpy(sd->sd_name, "CRYPTO", sizeof(sd->sd_name));
> > sd->sd_meta->ssdi.ssd_size = coerced_size;
> >
> > @@ -194,15 +194,12 @@ sr_crypto_assemble(struct sr_discipline
> > goto done;
> >
> > if (bc->bc_key_disk != NODEV) {
> > -
> > /* Read the mask key from the key disk. */
> > sd->mds.mdd_crypto.key_disk =
> > sr_crypto_read_key_disk(sd, bc->bc_key_disk);
> > if (sd->mds.mdd_crypto.key_disk == NULL)
> > goto done;
> > -
> > } else if (bc->bc_opaque_flags & BIOC_SOOUT) {
> > -
> > /* provide userland with kdf hint */
> > if (bc->bc_opaque == NULL)
> > goto done;
> > @@ -218,10 +215,8 @@ sr_crypto_assemble(struct sr_discipline
> > /* we're done */
> > bc->bc_opaque_status = BIOC_SOINOUT_OK;
> > rv = EAGAIN;
> > - goto done;
> > -
> > + goto done;
> > } else if (bc->bc_opaque_flags & BIOC_SOIN) {
> > -
> > /* get kdf with maskkey from userland */
> > if (sr_crypto_get_kdf(bc, sd))
> > goto done;
> > @@ -253,10 +248,11 @@ sr_crypto_getcryptop(struct sr_workunit
> > s = splbio();
> > uio = pool_get(&sd->mds.mdd_crypto.sr_uiopl, PR_ZERO | PR_NOWAIT);
> > if (uio == NULL)
> > - goto unwind;
> > - uio->uio_iov = pool_get(&sd->mds.mdd_crypto.sr_iovpl, PR_NOWAIT);
> > + goto poolunwind;
> > + uio->uio_iov = pool_get(&sd->mds.mdd_crypto.sr_iovpl,
> > + PR_ZERO | PR_NOWAIT);
> > if (uio->uio_iov == NULL)
> > - goto unwind;
> > + goto poolunwind;
> > splx(s);
> >
> > uio->uio_iovcnt = 1;
> > @@ -318,11 +314,14 @@ sr_crypto_getcryptop(struct sr_workunit
> > }
> >
> > return (crp);
> > +poolunwind:
> > + splx(s);
> > unwind:
> > if (crp)
> > crypto_freereq(crp);
> > if (uio && uio->uio_iov)
> > - if (wu->swu_xs->flags & SCSI_DATA_OUT)
> > + if ((wu->swu_xs->flags & SCSI_DATA_OUT) &&
> > + uio->uio_iov->iov_base)
> > free(uio->uio_iov->iov_base, M_DEVBUF);
> >
> > s = splbio();
> > @@ -346,7 +345,7 @@ sr_crypto_putcryptop(struct cryptop *crp
> > DNPRINTF(SR_D_DIS, "%s: sr_crypto_putcryptop crp: %p\n",
> > DEVNAME(wu->swu_dis->sd_sc), crp);
> >
> > - if (wu->swu_xs->flags & SCSI_DATA_OUT)
> > + if ((wu->swu_xs->flags & SCSI_DATA_OUT) && uio->uio_iov->iov_base)
> > free(uio->uio_iov->iov_base, M_DEVBUF);
> > s = splbio();
> > pool_put(&sd->mds.mdd_crypto.sr_iovpl, uio->uio_iov);
> > @@ -783,7 +782,7 @@ sr_crypto_create_key_disk(struct sr_disc
> > DEVNAME(sc), devname);
> > goto fail;
> > }
> > -
> > +
> > goto done;
> >
> > fail:
> > @@ -867,13 +866,7 @@ sr_crypto_read_key_disk(struct sr_discip
> > /*
> > * Read and validate key disk metadata.
> > */
> > - sm = malloc(SR_META_SIZE * 512, M_DEVBUF, M_NOWAIT|M_ZERO);
> > - if (sm == NULL) {
> > - printf("%s: not enough memory for metadata buffer\n",
> > - DEVNAME(sc));
> > - goto done;
> > - }
> > -
> > + sm = malloc(SR_META_SIZE * 512, M_DEVBUF, M_WAITOK | M_ZERO);
> > if (sr_meta_native_read(sd, dev, sm, NULL)) {
> > printf("%s: native bootprobe could not read native "
> > "metadata\n", DEVNAME(sc));
> > @@ -882,7 +875,7 @@ sr_crypto_read_key_disk(struct sr_discip
> >
> > if (sr_meta_validate(sd, dev, sm, NULL)) {
> > DNPRINTF(SR_D_META, "%s: invalid metadata\n",
> > - DEVNAME(sc));
> > + DEVNAME(sc));
> > goto done;
> > }
> >
> > @@ -893,13 +886,7 @@ sr_crypto_read_key_disk(struct sr_discip
> > }
> >
> > /* Construct key disk chunk. */
> > - key_disk = malloc(sizeof(struct sr_chunk), M_DEVBUF, M_NOWAIT|M_ZERO);
> > - if (key_disk == NULL) {
> > - printf("%s: not enough memory for chunk\n",
> > - DEVNAME(sc));
> > - goto done;
> > - }
> > -
> > + key_disk = malloc(sizeof(struct sr_chunk), M_DEVBUF, M_WAITOK | M_ZERO);
> > key_disk->src_dev_mm = dev;
> > key_disk->src_vn = vn;
> > key_disk->src_size = 0;
> > @@ -1106,8 +1093,8 @@ sr_crypto_meta_opt_load(struct sr_discip
> > sd->mds.mdd_crypto.scr_meta = &om->somi.som_meta.smm_crypto;
> > rv = 0;
> > }
> > -
> > - return rv;
> > +
> > + return (rv);
> > }
> >
> > int
> > @@ -1122,7 +1109,7 @@ sr_crypto_rw(struct sr_workunit *wu)
> > if (wu->swu_xs->flags & SCSI_DATA_OUT) {
> > crp = sr_crypto_getcryptop(wu, 1);
> > if (crp == NULL)
> > - panic("sr_crypto_getcryptop");
> > + panic("sr_crypto_rw: no crypto op");
> > crp->crp_callback = sr_crypto_write;
> > crp->crp_opaque = wu;
> > s = splvm();
> > @@ -1211,6 +1198,8 @@ sr_crypto_rw2(struct sr_workunit *wu, st
> >
> > LIST_INIT(&ccb->ccb_buf.b_dep);
> >
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_crypto_rw2", DEVNAME(sd->sd_sc));
> > TAILQ_INSERT_TAIL(&wu->swu_ccb, ccb, ccb_link);
> >
> > DNPRINTF(SR_D_DIS, "%s: %s: sr_crypto_rw2: b_bcount: %d "
> > @@ -1305,7 +1294,7 @@ sr_crypto_intr(struct buf *bp)
> > if ((xs->flags & SCSI_DATA_IN) && (xs->error == XS_NOERROR)) {
> > crp = sr_crypto_getcryptop(wu, 0);
> > if (crp == NULL)
> > - panic("sr_crypto_getcryptop");
> > + panic("sr_crypto_intr: no crypto op");
> > ccb->ccb_opaque = crp;
> > crp->crp_callback = sr_crypto_read;
> > crp->crp_opaque = wu;
> > @@ -1341,6 +1330,8 @@ sr_crypto_finish_io(struct sr_workunit *
> >
> > xs->resid = 0;
> >
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_crypto_finish_io", DEVNAME(sd->sd_sc));
> > TAILQ_FOREACH(ccb, &wu->swu_ccb, ccb_link) {
> > if (ccb->ccb_opaque == NULL)
> > continue;
> > Index: dev/softraid_raid1.c
> > ===================================================================
> > RCS file: /cvs/src/sys/dev/softraid_raid1.c,v
> > retrieving revision 1.25
> > diff -u -p -r1.25 softraid_raid1.c
> > --- dev/softraid_raid1.c 2 Jul 2010 09:20:26 -0000 1.25
> > +++ dev/softraid_raid1.c 27 Sep 2010 22:13:45 -0000
> > @@ -476,6 +476,8 @@ ragain:
> >
> > LIST_INIT(&b->b_dep);
> >
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_raid1_rw", DEVNAME(sd->sd_sc));
> > TAILQ_INSERT_TAIL(&wu->swu_ccb, ccb, ccb_link);
> >
> > DNPRINTF(SR_D_DIS, "%s: %s: sr_raid1: b_bcount: %d "
> > @@ -560,6 +562,9 @@ sr_raid1_intr(struct buf *bp)
> > printf("%s: retrying read on block %lld\n",
> > DEVNAME(sc), b->b_blkno);
> > sr_ccb_put(ccb);
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_raid1_intr_cb",
> > + DEVNAME(sd->sd_sc));
> > TAILQ_INIT(&wu->swu_ccb);
> > wu->swu_state = SR_WU_RESTART;
> > if (sd->sd_scsi_rw(wu))
> > @@ -647,6 +652,8 @@ sr_raid1_recreate_wu(struct sr_workunit
> > DNPRINTF(SR_D_INTR, "%s: sr_raid1_recreate_wu: %p\n", wup);
> >
> > /* toss all ccbs */
> > + if (wu->swu_cb_active == 1)
> > + panic("%s: sr_raid1_recreate_wu", DEVNAME(sd->sd_sc));
> > while ((ccb = TAILQ_FIRST(&wup->swu_ccb)) != NULL) {
> > TAILQ_REMOVE(&wup->swu_ccb, ccb, ccb_link);
> > sr_ccb_put(ccb);
> > Index: dev/softraidvar.h
> > ===================================================================
> > RCS file: /cvs/src/sys/dev/softraidvar.h,v
> > retrieving revision 1.95
> > diff -u -p -r1.95 softraidvar.h
> > --- dev/softraidvar.h 30 Aug 2010 17:32:40 -0000 1.95
> > +++ dev/softraidvar.h 26 Sep 2010 14:37:20 -0000
> > @@ -346,6 +346,11 @@ struct sr_workunit {
> > /* all ios that make up this workunit */
> > struct sr_ccb_list swu_ccb;
> >
> > + /* task memory */
> > + struct workq_task swu_wqt;
> > + struct workq_task swu_intr;
> > + int swu_cb_active; /* in callback */
> > +
> > TAILQ_ENTRY(sr_workunit) swu_link;
> > };
> >
> > @@ -484,6 +489,8 @@ struct sr_discipline {
> > #endif /* AOE */
> > } sd_dis_specific;/* dis specific members */
> > #define mds sd_dis_specific
> > +
> > + struct workq *sd_workq;
> >
> > /* discipline metadata */
> > struct sr_metadata *sd_meta; /* in memory copy of metadata */
