// SPDX-License-Identifier: GPL-2.0-only /* * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. * Copyright 2004-2011 Red Hat, Inc. */ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt #include #include #include #include #include #include #include #include "incore.h" #include "glock.h" #include "glops.h" #include "recovery.h" #include "util.h" #include "sys.h" #include "trace_gfs2.h" /** * gfs2_update_stats - Update time based stats * @s: The stats to update (local or global) * @index: The index inside @s * @sample: New data to include */ static inline void gfs2_update_stats(struct gfs2_lkstats *s, unsigned index, s64 sample) { /* * @delta is the difference between the current rtt sample and the * running average srtt. We add 1/8 of that to the srtt in order to * update the current srtt estimate. The variance estimate is a bit * more complicated. We subtract the current variance estimate from * the abs value of the @delta and add 1/4 of that to the running * total. That's equivalent to 3/4 of the current variance * estimate plus 1/4 of the abs of @delta. * * Note that the index points at the array entry containing the * smoothed mean value, and the variance is always in the following * entry * * Reference: TCP/IP Illustrated, vol 2, p. 831,832 * All times are in units of integer nanoseconds. Unlike the TCP/IP * case, they are not scaled fixed point. */ s64 delta = sample - s->stats[index]; s->stats[index] += (delta >> 3); index++; s->stats[index] += (s64)(abs(delta) - s->stats[index]) >> 2; } /** * gfs2_update_reply_times - Update locking statistics * @gl: The glock to update * * This assumes that gl->gl_dstamp has been set earlier. * * The rtt (lock round trip time) is an estimate of the time * taken to perform a dlm lock request. We update it on each * reply from the dlm. * * The blocking flag is set on the glock for all dlm requests * which may potentially block due to lock requests from other nodes. * DLM requests where the current lock state is exclusive, the * requested state is null (or unlocked) or where the TRY or * TRY_1CB flags are set are classified as non-blocking. All * other DLM requests are counted as (potentially) blocking. */ static inline void gfs2_update_reply_times(struct gfs2_glock *gl) { struct gfs2_pcpu_lkstats *lks; const unsigned gltype = gl->gl_name.ln_type; unsigned index = test_bit(GLF_BLOCKING, &gl->gl_flags) ? GFS2_LKS_SRTTB : GFS2_LKS_SRTT; s64 rtt; preempt_disable(); rtt = ktime_to_ns(ktime_sub(ktime_get_real(), gl->gl_dstamp)); lks = this_cpu_ptr(gl->gl_name.ln_sbd->sd_lkstats); gfs2_update_stats(&gl->gl_stats, index, rtt); /* Local */ gfs2_update_stats(&lks->lkstats[gltype], index, rtt); /* Global */ preempt_enable(); trace_gfs2_glock_lock_time(gl, rtt); } /** * gfs2_update_request_times - Update locking statistics * @gl: The glock to update * * The irt (lock inter-request times) measures the average time * between requests to the dlm. It is updated immediately before * each dlm call. */ static inline void gfs2_update_request_times(struct gfs2_glock *gl) { struct gfs2_pcpu_lkstats *lks; const unsigned gltype = gl->gl_name.ln_type; ktime_t dstamp; s64 irt; preempt_disable(); dstamp = gl->gl_dstamp; gl->gl_dstamp = ktime_get_real(); irt = ktime_to_ns(ktime_sub(gl->gl_dstamp, dstamp)); lks = this_cpu_ptr(gl->gl_name.ln_sbd->sd_lkstats); gfs2_update_stats(&gl->gl_stats, GFS2_LKS_SIRT, irt); /* Local */ gfs2_update_stats(&lks->lkstats[gltype], GFS2_LKS_SIRT, irt); /* Global */ preempt_enable(); } static void gdlm_ast(void *arg) { struct gfs2_glock *gl = arg; unsigned ret = gl->gl_state; gfs2_update_reply_times(gl); BUG_ON(gl->gl_lksb.sb_flags & DLM_SBF_DEMOTED); if ((gl->gl_lksb.sb_flags & DLM_SBF_VALNOTVALID) && gl->gl_lksb.sb_lvbptr) memset(gl->gl_lksb.sb_lvbptr, 0, GDLM_LVB_SIZE); switch (gl->gl_lksb.sb_status) { case -DLM_EUNLOCK: /* Unlocked, so glock can be freed */ if (gl->gl_ops->go_free) gl->gl_ops->go_free(gl); gfs2_glock_free(gl); return; case -DLM_ECANCEL: /* Cancel while getting lock */ ret |= LM_OUT_CANCELED; goto out; case -EAGAIN: /* Try lock fails */ case -EDEADLK: /* Deadlock detected */ goto out; case -ETIMEDOUT: /* Canceled due to timeout */ ret |= LM_OUT_ERROR; goto out; case 0: /* Success */ break; default: /* Something unexpected */ BUG(); } ret = gl->gl_req; if (gl->gl_lksb.sb_flags & DLM_SBF_ALTMODE) { if (gl->gl_req == LM_ST_SHARED) ret = LM_ST_DEFERRED; else if (gl->gl_req == LM_ST_DEFERRED) ret = LM_ST_SHARED; else BUG(); } set_bit(GLF_INITIAL, &gl->gl_flags); gfs2_glock_complete(gl, ret); return; out: if (!test_bit(GLF_INITIAL, &gl->gl_flags)) gl->gl_lksb.sb_lkid = 0; gfs2_glock_complete(gl, ret); } static void gdlm_bast(void *arg, int mode) { struct gfs2_glock *gl = arg; switch (mode) { case DLM_LOCK_EX: gfs2_glock_cb(gl, LM_ST_UNLOCKED); break; case DLM_LOCK_CW: gfs2_glock_cb(gl, LM_ST_DEFERRED); break; case DLM_LOCK_PR: gfs2_glock_cb(gl, LM_ST_SHARED); break; default: fs_err(gl->gl_name.ln_sbd, "unknown bast mode %d\n", mode); BUG(); } } /* convert gfs lock-state to dlm lock-mode */ static int make_mode(struct gfs2_sbd *sdp, const unsigned int lmstate) { switch (lmstate) { case LM_ST_UNLOCKED: return DLM_LOCK_NL; case LM_ST_EXCLUSIVE: return DLM_LOCK_EX; case LM_ST_DEFERRED: return DLM_LOCK_CW; case LM_ST_SHARED: return DLM_LOCK_PR; } fs_err(sdp, "unknown LM state %d\n", lmstate); BUG(); return -1; } static u32 make_flags(struct gfs2_glock *gl, const unsigned int gfs_flags, const int req) { u32 lkf = 0; if (gl->gl_lksb.sb_lvbptr) lkf |= DLM_LKF_VALBLK; if (gfs_flags & LM_FLAG_TRY) lkf |= DLM_LKF_NOQUEUE; if (gfs_flags & LM_FLAG_TRY_1CB) { lkf |= DLM_LKF_NOQUEUE; lkf |= DLM_LKF_NOQUEUEBAST; } if (gfs_flags & LM_FLAG_PRIORITY) { lkf |= DLM_LKF_NOORDER; lkf |= DLM_LKF_HEADQUE; } if (gfs_flags & LM_FLAG_ANY) { if (req == DLM_LOCK_PR) lkf |= DLM_LKF_ALTCW; else if (req == DLM_LOCK_CW) lkf |= DLM_LKF_ALTPR; else BUG(); } if (gl->gl_lksb.sb_lkid != 0) { lkf |= DLM_LKF_CONVERT; if (test_bit(GLF_BLOCKING, &gl->gl_flags)) lkf |= DLM_LKF_QUECVT; } return lkf; } static void gfs2_reverse_hex(char *c, u64 value) { *c = '0'; while (value) { *c-- = hex_asc[value & 0x0f]; value >>= 4; } } static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state, unsigned int flags) { struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct; int req; u32 lkf; char strname[GDLM_STRNAME_BYTES] = ""; int error; req = make_mode(gl->gl_name.ln_sbd, req_state); lkf = make_flags(gl, flags, req); gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT); gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT); if (gl->gl_lksb.sb_lkid) { gfs2_update_request_times(gl); } else { memset(strname, ' ', GDLM_STRNAME_BYTES - 1); strname[GDLM_STRNAME_BYTES - 1] = '\0'; gfs2_reverse_hex(strname + 7, gl->gl_name.ln_type); gfs2_reverse_hex(strname + 23, gl->gl_name.ln_number); gl->gl_dstamp = ktime_get_real(); } /* * Submit the actual lock request. */ again: error = dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, strname, GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast); if (error == -EBUSY) { msleep(20); goto again; } return error; } static void gdlm_put_lock(struct gfs2_glock *gl) { struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; struct lm_lockstruct *ls = &sdp->sd_lockstruct; int error; if (gl->gl_lksb.sb_lkid == 0) { gfs2_glock_free(gl); return; } clear_bit(GLF_BLOCKING, &gl->gl_flags); gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT); gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT); gfs2_update_request_times(gl); /* don't want to call dlm if we've unmounted the lock protocol */ if (test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) { gfs2_glock_free(gl); return; } /* don't want to skip dlm_unlock writing the lvb when lock has one */ if (test_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags) && !gl->gl_lksb.sb_lvbptr) { gfs2_glock_free(gl); return; } again: error = dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_VALBLK, NULL, gl); if (error == -EBUSY) { msleep(20); goto again; } if (error) { fs_err(sdp, "gdlm_unlock %x,%llx err=%d\n", gl->gl_name.ln_type, (unsigned long long)gl->gl_name.ln_number, error); return; } } static void gdlm_cancel(struct gfs2_glock *gl) { struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct; dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_CANCEL, NULL, gl); } /* * dlm/gfs2 recovery coordination using dlm_recover callbacks * * 0. gfs2 checks for another cluster node withdraw, needing journal replay * 1. dlm_controld sees lockspace members change * 2. dlm_controld blocks dlm-kernel locking activity * 3. dlm_controld within dlm-kernel notifies gfs2 (recover_prep) * 4. dlm_controld starts and finishes its own user level recovery * 5. dlm_controld starts dlm-kernel dlm_recoverd to do kernel recovery * 6. dlm_recoverd notifies gfs2 of failed nodes (recover_slot) * 7. dlm_recoverd does its own lock recovery * 8. dlm_recoverd unblocks dlm-kernel locking activity * 9. dlm_recoverd notifies gfs2 when done (recover_done with new generation) * 10. gfs2_control updates control_lock lvb with new generation and jid bits * 11. gfs2_control enqueues journals for gfs2_recover to recover (maybe none) * 12. gfs2_recover dequeues and recovers journals of failed nodes * 13. gfs2_recover provides recovery results to gfs2_control (recovery_result) * 14. gfs2_control updates control_lock lvb jid bits for recovered journals * 15. gfs2_control unblocks normal locking when all journals are recovered * * - failures during recovery * * recover_prep() may set BLOCK_LOCKS (step 3) again before gfs2_control * clears BLOCK_LOCKS (step 15), e.g. another node fails while still * recovering for a prior failure. gfs2_control needs a way to detect * this so it can leave BLOCK_LOCKS set in step 15. This is managed using * the recover_block and recover_start values. * * recover_done() provides a new lockspace generation number each time it * is called (step 9). This generation number is saved as recover_start. * When recover_prep() is called, it sets BLOCK_LOCKS and sets * recover_block = recover_start. So, while recover_block is equal to * recover_start, BLOCK_LOCKS should remain set. (recover_spin must * be held around the BLOCK_LOCKS/recover_block/recover_start logic.) * * - more specific gfs2 steps in sequence above * * 3. recover_prep sets BLOCK_LOCKS and sets recover_block = recover_start * 6. recover_slot records any failed jids (maybe none) * 9. recover_done sets recover_start = new generation number * 10. gfs2_control sets control_lock lvb = new gen + bits for failed jids * 12. gfs2_recover does journal recoveries for failed jids identified above * 14. gfs2_control clears control_lock lvb bits for recovered jids * 15. gfs2_control checks if recover_block == recover_start (step 3 occured * again) then do nothing, otherwise if recover_start > recover_block * then clear BLOCK_LOCKS. * * - parallel recovery steps across all nodes * * All nodes attempt to update the control_lock lvb with the new generation * number and jid bits, but only the first to get the control_lock EX will * do so; others will see that it's already done (lvb already contains new * generation number.) * * . All nodes get the same recover_prep/recover_slot/recover_done callbacks * . All nodes attempt to set control_lock lvb gen + bits for the new gen * . One node gets control_lock first and writes the lvb, others see it's done * . All nodes attempt to recover jids for which they see control_lock bits set * . One node succeeds for a jid, and that one clears the jid bit in the lvb * . All nodes will eventually see all lvb bits clear and unblock locks * * - is there a problem with clearing an lvb bit that should be set * and missing a journal recovery? * * 1. jid fails * 2. lvb bit set for step 1 * 3. jid recovered for step 1 * 4. jid taken again (new mount) * 5. jid fails (for step 4) * 6. lvb bit set for step 5 (will already be set) * 7. lvb bit cleared for step 3 * * This is not a problem because the failure in step 5 does not * require recovery, because the mount in step 4 could not have * progressed far enough to unblock locks and access the fs. The * control_mount() function waits for all recoveries to be complete * for the latest lockspace generation before ever unblocking locks * and returning. The mount in step 4 waits until the recovery in * step 1 is done. * * - special case of first mounter: first node to mount the fs * * The first node to mount a gfs2 fs needs to check all the journals * and recover any that need recovery before other nodes are allowed * to mount the fs. (Others may begin mounting, but they must wait * for the first mounter to be done before taking locks on the fs * or accessing the fs.) This has two parts: * * 1. The mounted_lock tells a node it's the first to mount the fs. * Each node holds the mounted_lock in PR while it's mounted. * Each node tries to acquire the mounted_lock in EX when it mounts. * If a node is granted the mounted_lock EX it means there are no * other mounted nodes (no PR locks exist), and it is the first mounter. * The mounted_lock is demoted to PR when first recovery is done, so * others will fail to get an EX lock, but will get a PR lock. * * 2. The control_lock blocks others in control_mount() while the first * mounter is doing first mount recovery of all journals. * A mounting node needs to acquire control_lock in EX mode before * it can proceed. The first mounter holds control_lock in EX while doing * the first mount recovery, blocking mounts from other nodes, then demotes * control_lock to NL when it's done (others_may_mount/first_done), * allowing other nodes to continue mounting. * * first mounter: * control_lock EX/NOQUEUE success * mounted_lock EX/NOQUEUE success (no other PR, so no other mounters) * set first=1 * do first mounter recovery * mounted_lock EX->PR * control_lock EX->NL, write lvb generation * * other mounter: * control_lock EX/NOQUEUE success (if fail -EAGAIN, retry) * mounted_lock EX/NOQUEUE fail -EAGAIN (expected due to other mounters PR) * mounted_lock PR/NOQUEUE success * read lvb generation * control_lock EX->NL * set first=0 * * - mount during recovery * * If a node mounts while others are doing recovery (not first mounter), * the mounting node will get its initial recover_done() callback without * having seen any previous failures/callbacks. * * It must wait for all recoveries preceding its mount to be finished * before it unblocks locks. It does this by repeating the "other mounter" * steps above until the lvb generation number is >= its mount generation * number (from initial recover_done) and all lvb bits are clear. * * - control_lock lvb format * * 4 bytes generation number: the latest dlm lockspace generation number * from recover_done callback. Indicates the jid bitmap has been updated * to reflect all slot failures through that generation. * 4 bytes unused. * GDLM_LVB_SIZE-8 bytes of jid bit map. If bit N is set, it indicates * that jid N needs recovery. */ #define JID_BITMAP_OFFSET 8 /* 4 byte generation number + 4 byte unused */ static void control_lvb_read(struct lm_lockstruct *ls, uint32_t *lvb_gen, char *lvb_bits) { __le32 gen; memcpy(lvb_bits, ls->ls_control_lvb, GDLM_LVB_SIZE); memcpy(&gen, lvb_bits, sizeof(__le32)); *lvb_gen = le32_to_cpu(gen); } static void control_lvb_write(struct lm_lockstruct *ls, uint32_t lvb_gen, char *lvb_bits) { __le32 gen; memcpy(ls->ls_control_lvb, lvb_bits, GDLM_LVB_SIZE); gen = cpu_to_le32(lvb_gen); memcpy(ls->ls_control_lvb, &gen, sizeof(__le32)); } static int all_jid_bits_clear(char *lvb) { return !memchr_inv(lvb + JID_BITMAP_OFFSET, 0, GDLM_LVB_SIZE - JID_BITMAP_OFFSET); } static void sync_wait_cb(void *arg) { struct lm_lockstruct *ls = arg; complete(&ls->ls_sync_wait); } static int sync_unlock(struct gfs2_sbd *sdp, struct dlm_lksb *lksb, char *name) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; int error; error = dlm_unlock(ls->ls_dlm, lksb->sb_lkid, 0, lksb, ls); if (error) { fs_err(sdp, "%s lkid %x error %d\n", name, lksb->sb_lkid, error); return error; } wait_for_completion(&ls->ls_sync_wait); if (lksb->sb_status != -DLM_EUNLOCK) { fs_err(sdp, "%s lkid %x status %d\n", name, lksb->sb_lkid, lksb->sb_status); return -1; } return 0; } static int sync_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags, unsigned int num, struct dlm_lksb *lksb, char *name) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; char strname[GDLM_STRNAME_BYTES]; int error, status; memset(strname, 0, GDLM_STRNAME_BYTES); snprintf(strname, GDLM_STRNAME_BYTES, "%8x%16x", LM_TYPE_NONDISK, num); error = dlm_lock(ls->ls_dlm, mode, lksb, flags, strname, GDLM_STRNAME_BYTES - 1, 0, sync_wait_cb, ls, NULL); if (error) { fs_err(sdp, "%s lkid %x flags %x mode %d error %d\n", name, lksb->sb_lkid, flags, mode, error); return error; } wait_for_completion(&ls->ls_sync_wait); status = lksb->sb_status; if (status && status != -EAGAIN) { fs_err(sdp, "%s lkid %x flags %x mode %d status %d\n", name, lksb->sb_lkid, flags, mode, status); } return status; } static int mounted_unlock(struct gfs2_sbd *sdp) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; return sync_unlock(sdp, &ls->ls_mounted_lksb, "mounted_lock"); } static int mounted_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; return sync_lock(sdp, mode, flags, GFS2_MOUNTED_LOCK, &ls->ls_mounted_lksb, "mounted_lock"); } static int control_unlock(struct gfs2_sbd *sdp) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; return sync_unlock(sdp, &ls->ls_control_lksb, "control_lock"); } static int control_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; return sync_lock(sdp, mode, flags, GFS2_CONTROL_LOCK, &ls->ls_control_lksb, "control_lock"); } /** * remote_withdraw - react to a node withdrawing from the file system * @sdp: The superblock */ static void remote_withdraw(struct gfs2_sbd *sdp) { struct gfs2_jdesc *jd; int ret = 0, count = 0; list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) continue; ret = gfs2_recover_journal(jd, true); if (ret) break; count++; } /* Now drop the additional reference we acquired */ fs_err(sdp, "Journals checked: %d, ret = %d.\n", count, ret); } static void gfs2_control_func(struct work_struct *work) { struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_control_work.work); struct lm_lockstruct *ls = &sdp->sd_lockstruct; uint32_t block_gen, start_gen, lvb_gen, flags; int recover_set = 0; int write_lvb = 0; int recover_size; int i, error; /* First check for other nodes that may have done a withdraw. */ if (test_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags)) { remote_withdraw(sdp); clear_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags); return; } spin_lock(&ls->ls_recover_spin); /* * No MOUNT_DONE means we're still mounting; control_mount() * will set this flag, after which this thread will take over * all further clearing of BLOCK_LOCKS. * * FIRST_MOUNT means this node is doing first mounter recovery, * for which recovery control is handled by * control_mount()/control_first_done(), not this thread. */ if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) || test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { spin_unlock(&ls->ls_recover_spin); return; } block_gen = ls->ls_recover_block; start_gen = ls->ls_recover_start; spin_unlock(&ls->ls_recover_spin); /* * Equal block_gen and start_gen implies we are between * recover_prep and recover_done callbacks, which means * dlm recovery is in progress and dlm locking is blocked. * There's no point trying to do any work until recover_done. */ if (block_gen == start_gen) return; /* * Propagate recover_submit[] and recover_result[] to lvb: * dlm_recoverd adds to recover_submit[] jids needing recovery * gfs2_recover adds to recover_result[] journal recovery results * * set lvb bit for jids in recover_submit[] if the lvb has not * yet been updated for the generation of the failure * * clear lvb bit for jids in recover_result[] if the result of * the journal recovery is SUCCESS */ error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_VALBLK); if (error) { fs_err(sdp, "control lock EX error %d\n", error); return; } control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits); spin_lock(&ls->ls_recover_spin); if (block_gen != ls->ls_recover_block || start_gen != ls->ls_recover_start) { fs_info(sdp, "recover generation %u block1 %u %u\n", start_gen, block_gen, ls->ls_recover_block); spin_unlock(&ls->ls_recover_spin); control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT); return; } recover_size = ls->ls_recover_size; if (lvb_gen <= start_gen) { /* * Clear lvb bits for jids we've successfully recovered. * Because all nodes attempt to recover failed journals, * a journal can be recovered multiple times successfully * in succession. Only the first will really do recovery, * the others find it clean, but still report a successful * recovery. So, another node may have already recovered * the jid and cleared the lvb bit for it. */ for (i = 0; i < recover_size; i++) { if (ls->ls_recover_result[i] != LM_RD_SUCCESS) continue; ls->ls_recover_result[i] = 0; if (!test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET)) continue; __clear_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET); write_lvb = 1; } } if (lvb_gen == start_gen) { /* * Failed slots before start_gen are already set in lvb. */ for (i = 0; i < recover_size; i++) { if (!ls->ls_recover_submit[i]) continue; if (ls->ls_recover_submit[i] < lvb_gen) ls->ls_recover_submit[i] = 0; } } else if (lvb_gen < start_gen) { /* * Failed slots before start_gen are not yet set in lvb. */ for (i = 0; i < recover_size; i++) { if (!ls->ls_recover_submit[i]) continue; if (ls->ls_recover_submit[i] < start_gen) { ls->ls_recover_submit[i] = 0; __set_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET); } } /* even if there are no bits to set, we need to write the latest generation to the lvb */ write_lvb = 1; } else { /* * we should be getting a recover_done() for lvb_gen soon */ } spin_unlock(&ls->ls_recover_spin); if (write_lvb) { control_lvb_write(ls, start_gen, ls->ls_lvb_bits); flags = DLM_LKF_CONVERT | DLM_LKF_VALBLK; } else { flags = DLM_LKF_CONVERT; } error = control_lock(sdp, DLM_LOCK_NL, flags); if (error) { fs_err(sdp, "control lock NL error %d\n", error); return; } /* * Everyone will see jid bits set in the lvb, run gfs2_recover_set(), * and clear a jid bit in the lvb if the recovery is a success. * Eventually all journals will be recovered, all jid bits will * be cleared in the lvb, and everyone will clear BLOCK_LOCKS. */ for (i = 0; i < recover_size; i++) { if (test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET)) { fs_info(sdp, "recover generation %u jid %d\n", start_gen, i); gfs2_recover_set(sdp, i); recover_set++; } } if (recover_set) return; /* * No more jid bits set in lvb, all recovery is done, unblock locks * (unless a new recover_prep callback has occured blocking locks * again while working above) */ spin_lock(&ls->ls_recover_spin); if (ls->ls_recover_block == block_gen && ls->ls_recover_start == start_gen) { clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); fs_info(sdp, "recover generation %u done\n", start_gen); gfs2_glock_thaw(sdp); } else { fs_info(sdp, "recover generation %u block2 %u %u\n", start_gen, block_gen, ls->ls_recover_block); spin_unlock(&ls->ls_recover_spin); } } static int control_mount(struct gfs2_sbd *sdp) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; uint32_t start_gen, block_gen, mount_gen, lvb_gen; int mounted_mode; int retries = 0; int error; memset(&ls->ls_mounted_lksb, 0, sizeof(struct dlm_lksb)); memset(&ls->ls_control_lksb, 0, sizeof(struct dlm_lksb)); memset(&ls->ls_control_lvb, 0, GDLM_LVB_SIZE); ls->ls_control_lksb.sb_lvbptr = ls->ls_control_lvb; init_completion(&ls->ls_sync_wait); set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags); error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_VALBLK); if (error) { fs_err(sdp, "control_mount control_lock NL error %d\n", error); return error; } error = mounted_lock(sdp, DLM_LOCK_NL, 0); if (error) { fs_err(sdp, "control_mount mounted_lock NL error %d\n", error); control_unlock(sdp); return error; } mounted_mode = DLM_LOCK_NL; restart: if (retries++ && signal_pending(current)) { error = -EINTR; goto fail; } /* * We always start with both locks in NL. control_lock is * demoted to NL below so we don't need to do it here. */ if (mounted_mode != DLM_LOCK_NL) { error = mounted_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT); if (error) goto fail; mounted_mode = DLM_LOCK_NL; } /* * Other nodes need to do some work in dlm recovery and gfs2_control * before the recover_done and control_lock will be ready for us below. * A delay here is not required but often avoids having to retry. */ msleep_interruptible(500); /* * Acquire control_lock in EX and mounted_lock in either EX or PR. * control_lock lvb keeps track of any pending journal recoveries. * mounted_lock indicates if any other nodes have the fs mounted. */ error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE|DLM_LKF_VALBLK); if (error == -EAGAIN) { goto restart; } else if (error) { fs_err(sdp, "control_mount control_lock EX error %d\n", error); goto fail; } /** * If we're a spectator, we don't want to take the lock in EX because * we cannot do the first-mount responsibility it implies: recovery. */ if (sdp->sd_args.ar_spectator) goto locks_done; error = mounted_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE); if (!error) { mounted_mode = DLM_LOCK_EX; goto locks_done; } else if (error != -EAGAIN) { fs_err(sdp, "control_mount mounted_lock EX error %d\n", error); goto fail; } error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE); if (!error) { mounted_mode = DLM_LOCK_PR; goto locks_done; } else { /* not even -EAGAIN should happen here */ fs_err(sdp, "control_mount mounted_lock PR error %d\n", error); goto fail; } locks_done: /* * If we got both locks above in EX, then we're the first mounter. * If not, then we need to wait for the control_lock lvb to be * updated by other mounted nodes to reflect our mount generation. * * In simple first mounter cases, first mounter will see zero lvb_gen, * but in cases where all existing nodes leave/fail before mounting * nodes finish control_mount, then all nodes will be mounting and * lvb_gen will be non-zero. */ control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits); if (lvb_gen == 0xFFFFFFFF) { /* special value to force mount attempts to fail */ fs_err(sdp, "control_mount control_lock disabled\n"); error = -EINVAL; goto fail; } if (mounted_mode == DLM_LOCK_EX) { /* first mounter, keep both EX while doing first recovery */ spin_lock(&ls->ls_recover_spin); clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags); set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags); set_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); fs_info(sdp, "first mounter control generation %u\n", lvb_gen); return 0; } error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT); if (error) goto fail; /* * We are not first mounter, now we need to wait for the control_lock * lvb generation to be >= the generation from our first recover_done * and all lvb bits to be clear (no pending journal recoveries.) */ if (!all_jid_bits_clear(ls->ls_lvb_bits)) { /* journals need recovery, wait until all are clear */ fs_info(sdp, "control_mount wait for journal recovery\n"); goto restart; } spin_lock(&ls->ls_recover_spin); block_gen = ls->ls_recover_block; start_gen = ls->ls_recover_start; mount_gen = ls->ls_recover_mount; if (lvb_gen < mount_gen) { /* wait for mounted nodes to update control_lock lvb to our generation, which might include new recovery bits set */ if (sdp->sd_args.ar_spectator) { fs_info(sdp, "Recovery is required. Waiting for a " "non-spectator to mount.\n"); msleep_interruptible(1000); } else { fs_info(sdp, "control_mount wait1 block %u start %u " "mount %u lvb %u flags %lx\n", block_gen, start_gen, mount_gen, lvb_gen, ls->ls_recover_flags); } spin_unlock(&ls->ls_recover_spin); goto restart; } if (lvb_gen != start_gen) { /* wait for mounted nodes to update control_lock lvb to the latest recovery generation */ fs_info(sdp, "control_mount wait2 block %u start %u mount %u " "lvb %u flags %lx\n", block_gen, start_gen, mount_gen, lvb_gen, ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); goto restart; } if (block_gen == start_gen) { /* dlm recovery in progress, wait for it to finish */ fs_info(sdp, "control_mount wait3 block %u start %u mount %u " "lvb %u flags %lx\n", block_gen, start_gen, mount_gen, lvb_gen, ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); goto restart; } clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags); set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags); memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t)); memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t)); spin_unlock(&ls->ls_recover_spin); return 0; fail: mounted_unlock(sdp); control_unlock(sdp); return error; } static int control_first_done(struct gfs2_sbd *sdp) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; uint32_t start_gen, block_gen; int error; restart: spin_lock(&ls->ls_recover_spin); start_gen = ls->ls_recover_start; block_gen = ls->ls_recover_block; if (test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags) || !test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) || !test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { /* sanity check, should not happen */ fs_err(sdp, "control_first_done start %u block %u flags %lx\n", start_gen, block_gen, ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); control_unlock(sdp); return -1; } if (start_gen == block_gen) { /* * Wait for the end of a dlm recovery cycle to switch from * first mounter recovery. We can ignore any recover_slot * callbacks between the recover_prep and next recover_done * because we are still the first mounter and any failed nodes * have not fully mounted, so they don't need recovery. */ spin_unlock(&ls->ls_recover_spin); fs_info(sdp, "control_first_done wait gen %u\n", start_gen); wait_on_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY, TASK_UNINTERRUPTIBLE); goto restart; } clear_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags); set_bit(DFL_FIRST_MOUNT_DONE, &ls->ls_recover_flags); memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t)); memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t)); spin_unlock(&ls->ls_recover_spin); memset(ls->ls_lvb_bits, 0, GDLM_LVB_SIZE); control_lvb_write(ls, start_gen, ls->ls_lvb_bits); error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT); if (error) fs_err(sdp, "control_first_done mounted PR error %d\n", error); error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT|DLM_LKF_VALBLK); if (error) fs_err(sdp, "control_first_done control NL error %d\n", error); return error; } /* * Expand static jid arrays if necessary (by increments of RECOVER_SIZE_INC) * to accomodate the largest slot number. (NB dlm slot numbers start at 1, * gfs2 jids start at 0, so jid = slot - 1) */ #define RECOVER_SIZE_INC 16 static int set_recover_size(struct gfs2_sbd *sdp, struct dlm_slot *slots, int num_slots) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; uint32_t *submit = NULL; uint32_t *result = NULL; uint32_t old_size, new_size; int i, max_jid; if (!ls->ls_lvb_bits) { ls->ls_lvb_bits = kzalloc(GDLM_LVB_SIZE, GFP_NOFS); if (!ls->ls_lvb_bits) return -ENOMEM; } max_jid = 0; for (i = 0; i < num_slots; i++) { if (max_jid < slots[i].slot - 1) max_jid = slots[i].slot - 1; } old_size = ls->ls_recover_size; new_size = old_size; while (new_size < max_jid + 1) new_size += RECOVER_SIZE_INC; if (new_size == old_size) return 0; submit = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS); result = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS); if (!submit || !result) { kfree(submit); kfree(result); return -ENOMEM; } spin_lock(&ls->ls_recover_spin); memcpy(submit, ls->ls_recover_submit, old_size * sizeof(uint32_t)); memcpy(result, ls->ls_recover_result, old_size * sizeof(uint32_t)); kfree(ls->ls_recover_submit); kfree(ls->ls_recover_result); ls->ls_recover_submit = submit; ls->ls_recover_result = result; ls->ls_recover_size = new_size; spin_unlock(&ls->ls_recover_spin); return 0; } static void free_recover_size(struct lm_lockstruct *ls) { kfree(ls->ls_lvb_bits); kfree(ls->ls_recover_submit); kfree(ls->ls_recover_result); ls->ls_recover_submit = NULL; ls->ls_recover_result = NULL; ls->ls_recover_size = 0; ls->ls_lvb_bits = NULL; } /* dlm calls before it does lock recovery */ static void gdlm_recover_prep(void *arg) { struct gfs2_sbd *sdp = arg; struct lm_lockstruct *ls = &sdp->sd_lockstruct; if (gfs2_withdrawn(sdp)) { fs_err(sdp, "recover_prep ignored due to withdraw.\n"); return; } spin_lock(&ls->ls_recover_spin); ls->ls_recover_block = ls->ls_recover_start; set_bit(DFL_DLM_RECOVERY, &ls->ls_recover_flags); if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) || test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { spin_unlock(&ls->ls_recover_spin); return; } set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); } /* dlm calls after recover_prep has been completed on all lockspace members; identifies slot/jid of failed member */ static void gdlm_recover_slot(void *arg, struct dlm_slot *slot) { struct gfs2_sbd *sdp = arg; struct lm_lockstruct *ls = &sdp->sd_lockstruct; int jid = slot->slot - 1; if (gfs2_withdrawn(sdp)) { fs_err(sdp, "recover_slot jid %d ignored due to withdraw.\n", jid); return; } spin_lock(&ls->ls_recover_spin); if (ls->ls_recover_size < jid + 1) { fs_err(sdp, "recover_slot jid %d gen %u short size %d\n", jid, ls->ls_recover_block, ls->ls_recover_size); spin_unlock(&ls->ls_recover_spin); return; } if (ls->ls_recover_submit[jid]) { fs_info(sdp, "recover_slot jid %d gen %u prev %u\n", jid, ls->ls_recover_block, ls->ls_recover_submit[jid]); } ls->ls_recover_submit[jid] = ls->ls_recover_block; spin_unlock(&ls->ls_recover_spin); } /* dlm calls after recover_slot and after it completes lock recovery */ static void gdlm_recover_done(void *arg, struct dlm_slot *slots, int num_slots, int our_slot, uint32_t generation) { struct gfs2_sbd *sdp = arg; struct lm_lockstruct *ls = &sdp->sd_lockstruct; if (gfs2_withdrawn(sdp)) { fs_err(sdp, "recover_done ignored due to withdraw.\n"); return; } /* ensure the ls jid arrays are large enough */ set_recover_size(sdp, slots, num_slots); spin_lock(&ls->ls_recover_spin); ls->ls_recover_start = generation; if (!ls->ls_recover_mount) { ls->ls_recover_mount = generation; ls->ls_jid = our_slot - 1; } if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work, 0); clear_bit(DFL_DLM_RECOVERY, &ls->ls_recover_flags); smp_mb__after_atomic(); wake_up_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY); spin_unlock(&ls->ls_recover_spin); } /* gfs2_recover thread has a journal recovery result */ static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid, unsigned int result) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; if (gfs2_withdrawn(sdp)) { fs_err(sdp, "recovery_result jid %d ignored due to withdraw.\n", jid); return; } if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags)) return; /* don't care about the recovery of own journal during mount */ if (jid == ls->ls_jid) return; spin_lock(&ls->ls_recover_spin); if (test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { spin_unlock(&ls->ls_recover_spin); return; } if (ls->ls_recover_size < jid + 1) { fs_err(sdp, "recovery_result jid %d short size %d\n", jid, ls->ls_recover_size); spin_unlock(&ls->ls_recover_spin); return; } fs_info(sdp, "recover jid %d result %s\n", jid, result == LM_RD_GAVEUP ? "busy" : "success"); ls->ls_recover_result[jid] = result; /* GAVEUP means another node is recovering the journal; delay our next attempt to recover it, to give the other node a chance to finish before trying again */ if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work, result == LM_RD_GAVEUP ? HZ : 0); spin_unlock(&ls->ls_recover_spin); } static const struct dlm_lockspace_ops gdlm_lockspace_ops = { .recover_prep = gdlm_recover_prep, .recover_slot = gdlm_recover_slot, .recover_done = gdlm_recover_done, }; static int gdlm_mount(struct gfs2_sbd *sdp, const char *table) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; char cluster[GFS2_LOCKNAME_LEN]; const char *fsname; uint32_t flags; int error, ops_result; /* * initialize everything */ INIT_DELAYED_WORK(&sdp->sd_control_work, gfs2_control_func); spin_lock_init(&ls->ls_recover_spin); ls->ls_recover_flags = 0; ls->ls_recover_mount = 0; ls->ls_recover_start = 0; ls->ls_recover_block = 0; ls->ls_recover_size = 0; ls->ls_recover_submit = NULL; ls->ls_recover_result = NULL; ls->ls_lvb_bits = NULL; error = set_recover_size(sdp, NULL, 0); if (error) goto fail; /* * prepare dlm_new_lockspace args */ fsname = strchr(table, ':'); if (!fsname) { fs_info(sdp, "no fsname found\n"); error = -EINVAL; goto fail_free; } memset(cluster, 0, sizeof(cluster)); memcpy(cluster, table, strlen(table) - strlen(fsname)); fsname++; flags = DLM_LSFL_FS | DLM_LSFL_NEWEXCL; /* * create/join lockspace */ error = dlm_new_lockspace(fsname, cluster, flags, GDLM_LVB_SIZE, &gdlm_lockspace_ops, sdp, &ops_result, &ls->ls_dlm); if (error) { fs_err(sdp, "dlm_new_lockspace error %d\n", error); goto fail_free; } if (ops_result < 0) { /* * dlm does not support ops callbacks, * old dlm_controld/gfs_controld are used, try without ops. */ fs_info(sdp, "dlm lockspace ops not used\n"); free_recover_size(ls); set_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags); return 0; } if (!test_bit(SDF_NOJOURNALID, &sdp->sd_flags)) { fs_err(sdp, "dlm lockspace ops disallow jid preset\n"); error = -EINVAL; goto fail_release; } /* * control_mount() uses control_lock to determine first mounter, * and for later mounts, waits for any recoveries to be cleared. */ error = control_mount(sdp); if (error) { fs_err(sdp, "mount control error %d\n", error); goto fail_release; } ls->ls_first = !!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags); clear_bit(SDF_NOJOURNALID, &sdp->sd_flags); smp_mb__after_atomic(); wake_up_bit(&sdp->sd_flags, SDF_NOJOURNALID); return 0; fail_release: dlm_release_lockspace(ls->ls_dlm, 2); fail_free: free_recover_size(ls); fail: return error; } static void gdlm_first_done(struct gfs2_sbd *sdp) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; int error; if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags)) return; error = control_first_done(sdp); if (error) fs_err(sdp, "mount first_done error %d\n", error); } static void gdlm_unmount(struct gfs2_sbd *sdp) { struct lm_lockstruct *ls = &sdp->sd_lockstruct; if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags)) goto release; /* wait for gfs2_control_wq to be done with this mount */ spin_lock(&ls->ls_recover_spin); set_bit(DFL_UNMOUNT, &ls->ls_recover_flags); spin_unlock(&ls->ls_recover_spin); flush_delayed_work(&sdp->sd_control_work); /* mounted_lock and control_lock will be purged in dlm recovery */ release: if (ls->ls_dlm) { dlm_release_lockspace(ls->ls_dlm, 2); ls->ls_dlm = NULL; } free_recover_size(ls); } static const match_table_t dlm_tokens = { { Opt_jid, "jid=%d"}, { Opt_id, "id=%d"}, { Opt_first, "first=%d"}, { Opt_nodir, "nodir=%d"}, { Opt_err, NULL }, }; const struct lm_lockops gfs2_dlm_ops = { .lm_proto_name = "lock_dlm", .lm_mount = gdlm_mount, .lm_first_done = gdlm_first_done, .lm_recovery_result = gdlm_recovery_result, .lm_unmount = gdlm_unmount, .lm_put_lock = gdlm_put_lock, .lm_lock = gdlm_lock, .lm_cancel = gdlm_cancel, .lm_tokens = &dlm_tokens, }; ='#n1110'>1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 
// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Janz MODULbus VMOD-ICAN3 CAN Interface Driver
 *
 * Copyright (c) 2010 Ira W. Snyder <iws@ovro.caltech.edu>
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
#include <linux/platform_device.h>

#include <linux/netdevice.h>
#include <linux/can.h>
#include <linux/can/dev.h>
#include <linux/can/skb.h>
#include <linux/can/error.h>

#include <linux/mfd/janz.h>
#include <asm/io.h>

/* the DPM has 64k of memory, organized into 256x 256 byte pages */
#define DPM_NUM_PAGES		256
#define DPM_PAGE_SIZE		256
#define DPM_PAGE_ADDR(p)	((p) * DPM_PAGE_SIZE)

/* JANZ ICAN3 "old-style" host interface queue page numbers */
#define QUEUE_OLD_CONTROL	0
#define QUEUE_OLD_RB0		1
#define QUEUE_OLD_RB1		2
#define QUEUE_OLD_WB0		3
#define QUEUE_OLD_WB1		4

/* Janz ICAN3 "old-style" host interface control registers */
#define MSYNC_PEER		0x00		/* ICAN only */
#define MSYNC_LOCL		0x01		/* host only */
#define TARGET_RUNNING		0x02
#define FIRMWARE_STAMP		0x60		/* big endian firmware stamp */

#define MSYNC_RB0		0x01
#define MSYNC_RB1		0x02
#define MSYNC_RBLW		0x04
#define MSYNC_RB_MASK		(MSYNC_RB0 | MSYNC_RB1)

#define MSYNC_WB0		0x10
#define MSYNC_WB1		0x20
#define MSYNC_WBLW		0x40
#define MSYNC_WB_MASK		(MSYNC_WB0 | MSYNC_WB1)

/* Janz ICAN3 "new-style" host interface queue page numbers */
#define QUEUE_TOHOST		5
#define QUEUE_FROMHOST_MID	6
#define QUEUE_FROMHOST_HIGH	7
#define QUEUE_FROMHOST_LOW	8

/* The first free page in the DPM is #9 */
#define DPM_FREE_START		9

/* Janz ICAN3 "new-style" and "fast" host interface descriptor flags */
#define DESC_VALID		0x80
#define DESC_WRAP		0x40
#define DESC_INTERRUPT		0x20
#define DESC_IVALID		0x10
#define DESC_LEN(len)		(len)

/* Janz ICAN3 Firmware Messages */
#define MSG_CONNECTI		0x02
#define MSG_DISCONNECT		0x03
#define MSG_IDVERS		0x04
#define MSG_MSGLOST		0x05
#define MSG_NEWHOSTIF		0x08
#define MSG_INQUIRY		0x0a
#define MSG_SETAFILMASK		0x10
#define MSG_INITFDPMQUEUE	0x11
#define MSG_HWCONF		0x12
#define MSG_FMSGLOST		0x15
#define MSG_CEVTIND		0x37
#define MSG_CBTRREQ		0x41
#define MSG_COFFREQ		0x42
#define MSG_CONREQ		0x43
#define MSG_CCONFREQ		0x47
#define MSG_NMTS		0xb0
#define MSG_LMTS		0xb4

/*
 * Janz ICAN3 CAN Inquiry Message Types
 *
 * NOTE: there appears to be a firmware bug here. You must send
 * NOTE: INQUIRY_STATUS and expect to receive an INQUIRY_EXTENDED
 * NOTE: response. The controller never responds to a message with
 * NOTE: the INQUIRY_EXTENDED subspec :(
 */
#define INQUIRY_STATUS		0x00
#define INQUIRY_TERMINATION	0x01
#define INQUIRY_EXTENDED	0x04

/* Janz ICAN3 CAN Set Acceptance Filter Mask Message Types */
#define SETAFILMASK_REJECT	0x00
#define SETAFILMASK_FASTIF	0x02

/* Janz ICAN3 CAN Hardware Configuration Message Types */
#define HWCONF_TERMINATE_ON	0x01
#define HWCONF_TERMINATE_OFF	0x00

/* Janz ICAN3 CAN Event Indication Message Types */
#define CEVTIND_EI		0x01
#define CEVTIND_DOI		0x02
#define CEVTIND_LOST		0x04
#define CEVTIND_FULL		0x08
#define CEVTIND_BEI		0x10

#define CEVTIND_CHIP_SJA1000	0x02

#define ICAN3_BUSERR_QUOTA_MAX	255

/* Janz ICAN3 CAN Frame Conversion */
#define ICAN3_SNGL	0x02
#define ICAN3_ECHO	0x10
#define ICAN3_EFF_RTR	0x40
#define ICAN3_SFF_RTR	0x10
#define ICAN3_EFF	0x80

#define ICAN3_CAN_TYPE_MASK	0x0f
#define ICAN3_CAN_TYPE_SFF	0x00
#define ICAN3_CAN_TYPE_EFF	0x01

#define ICAN3_CAN_DLC_MASK	0x0f

/* Janz ICAN3 NMTS subtypes */
#define NMTS_CREATE_NODE_REQ	0x0
#define NMTS_SLAVE_STATE_IND	0x8
#define NMTS_SLAVE_EVENT_IND	0x9

/* Janz ICAN3 LMTS subtypes */
#define LMTS_BUSON_REQ		0x0
#define LMTS_BUSOFF_REQ		0x1
#define LMTS_CAN_CONF_REQ	0x2

/* Janz ICAN3 NMTS Event indications */
#define NE_LOCAL_OCCURRED	0x3
#define NE_LOCAL_RESOLVED	0x2
#define NE_REMOTE_OCCURRED	0xc
#define NE_REMOTE_RESOLVED	0x8

/*
 * SJA1000 Status and Error Register Definitions
 *
 * Copied from drivers/net/can/sja1000/sja1000.h
 */

/* status register content */
#define SR_BS		0x80
#define SR_ES		0x40
#define SR_TS		0x20
#define SR_RS		0x10
#define SR_TCS		0x08
#define SR_TBS		0x04
#define SR_DOS		0x02
#define SR_RBS		0x01

#define SR_CRIT (SR_BS|SR_ES)

/* ECC register */
#define ECC_SEG		0x1F
#define ECC_DIR		0x20
#define ECC_ERR		6
#define ECC_BIT		0x00
#define ECC_FORM	0x40
#define ECC_STUFF	0x80
#define ECC_MASK	0xc0

/* Number of buffers for use in the "new-style" host interface */
#define ICAN3_NEW_BUFFERS	16

/* Number of buffers for use in the "fast" host interface */
#define ICAN3_TX_BUFFERS	512
#define ICAN3_RX_BUFFERS	1024

/* SJA1000 Clock Input */
#define ICAN3_CAN_CLOCK		8000000

/* Janz ICAN3 firmware types */
enum ican3_fwtype {
	ICAN3_FWTYPE_ICANOS,
	ICAN3_FWTYPE_CAL_CANOPEN,
};

/* Driver Name */
#define DRV_NAME "janz-ican3"

/* DPM Control Registers -- starts at offset 0x100 in the MODULbus registers */
struct ican3_dpm_control {
	/* window address register */
	u8 window_address;
	u8 unused1;

	/*
	 * Read access: clear interrupt from microcontroller
	 * Write access: send interrupt to microcontroller
	 */
	u8 interrupt;
	u8 unused2;

	/* write-only: reset all hardware on the module */
	u8 hwreset;
	u8 unused3;

	/* write-only: generate an interrupt to the TPU */
	u8 tpuinterrupt;
};

struct ican3_dev {

	/* must be the first member */
	struct can_priv can;

	/* CAN network device */
	struct net_device *ndev;
	struct napi_struct napi;

	/* module number */
	unsigned int num;

	/* base address of registers and IRQ */
	struct janz_cmodio_onboard_regs __iomem *ctrl;
	struct ican3_dpm_control __iomem *dpmctrl;
	void __iomem *dpm;
	int irq;

	/* CAN bus termination status */
	struct completion termination_comp;
	bool termination_enabled;

	/* CAN bus error status registers */
	struct completion buserror_comp;
	struct can_berr_counter bec;

	/* firmware type */
	enum ican3_fwtype fwtype;
	char fwinfo[32];

	/* old and new style host interface */
	unsigned int iftype;

	/* queue for echo packets */
	struct sk_buff_head echoq;

	/*
	 * Any function which changes the current DPM page must hold this
	 * lock while it is performing data accesses. This ensures that the
	 * function will not be preempted and end up reading data from a
	 * different DPM page than it expects.
	 */
	spinlock_t lock;

	/* new host interface */
	unsigned int rx_int;
	unsigned int rx_num;
	unsigned int tx_num;

	/* fast host interface */
	unsigned int fastrx_start;
	unsigned int fastrx_num;
	unsigned int fasttx_start;
	unsigned int fasttx_num;

	/* first free DPM page */
	unsigned int free_page;
};

struct ican3_msg {
	u8 control;
	u8 spec;
	__le16 len;
	u8 data[252];
};

struct ican3_new_desc {
	u8 control;
	u8 pointer;
};

struct ican3_fast_desc {
	u8 control;
	u8 command;
	u8 data[14];
};

/* write to the window basic address register */
static inline void ican3_set_page(struct ican3_dev *mod, unsigned int page)
{
	BUG_ON(page >= DPM_NUM_PAGES);
	iowrite8(page, &mod->dpmctrl->window_address);
}

/*
 * ICAN3 "old-style" host interface
 */

/*
 * Receive a message from the ICAN3 "old-style" firmware interface
 *
 * LOCKING: must hold mod->lock
 *
 * returns 0 on success, -ENOMEM when no message exists
 */
static int ican3_old_recv_msg(struct ican3_dev *mod, struct ican3_msg *msg)
{
	unsigned int mbox, mbox_page;
	u8 locl, peer, xord;

	/* get the MSYNC registers */
	ican3_set_page(mod, QUEUE_OLD_CONTROL);
	peer = ioread8(mod->dpm + MSYNC_PEER);
	locl = ioread8(mod->dpm + MSYNC_LOCL);
	xord = locl ^ peer;

	if ((xord & MSYNC_RB_MASK) == 0x00) {
		netdev_dbg(mod->ndev, "no mbox for reading\n");
		return -ENOMEM;
	}

	/* find the first free mbox to read */
	if ((xord & MSYNC_RB_MASK) == MSYNC_RB_MASK)
		mbox = (xord & MSYNC_RBLW) ? MSYNC_RB0 : MSYNC_RB1;
	else
		mbox = (xord & MSYNC_RB0) ? MSYNC_RB0 : MSYNC_RB1;

	/* copy the message */
	mbox_page = (mbox == MSYNC_RB0) ? QUEUE_OLD_RB0 : QUEUE_OLD_RB1;
	ican3_set_page(mod, mbox_page);
	memcpy_fromio(msg, mod->dpm, sizeof(*msg));

	/*
	 * notify the firmware that the read buffer is available
	 * for it to fill again
	 */
	locl ^= mbox;

	ican3_set_page(mod, QUEUE_OLD_CONTROL);
	iowrite8(locl, mod->dpm + MSYNC_LOCL);
	return 0;
}

/*
 * Send a message through the "old-style" firmware interface
 *
 * LOCKING: must hold mod->lock
 *
 * returns 0 on success, -ENOMEM when no free space exists
 */
static int ican3_old_send_msg(struct ican3_dev *mod, struct ican3_msg *msg)
{
	unsigned int mbox, mbox_page;
	u8 locl, peer, xord;

	/* get the MSYNC registers */
	ican3_set_page(mod, QUEUE_OLD_CONTROL);
	peer = ioread8(mod->dpm + MSYNC_PEER);
	locl = ioread8(mod->dpm + MSYNC_LOCL);
	xord = locl ^ peer;

	if ((xord & MSYNC_WB_MASK) == MSYNC_WB_MASK) {
		netdev_err(mod->ndev, "no mbox for writing\n");
		return -ENOMEM;
	}

	/* calculate a free mbox to use */
	mbox = (xord & MSYNC_WB0) ? MSYNC_WB1 : MSYNC_WB0;

	/* copy the message to the DPM */
	mbox_page = (mbox == MSYNC_WB0) ? QUEUE_OLD_WB0 : QUEUE_OLD_WB1;
	ican3_set_page(mod, mbox_page);
	memcpy_toio(mod->dpm, msg, sizeof(*msg));

	locl ^= mbox;
	if (mbox == MSYNC_WB1)
		locl |= MSYNC_WBLW;

	ican3_set_page(mod, QUEUE_OLD_CONTROL);
	iowrite8(locl, mod->dpm + MSYNC_LOCL);
	return 0;
}

/*
 * ICAN3 "new-style" Host Interface Setup
 */

static void ican3_init_new_host_interface(struct ican3_dev *mod)
{
	struct ican3_new_desc desc;
	unsigned long flags;
	void __iomem *dst;
	int i;

	spin_lock_irqsave(&mod->lock, flags);

	/* setup the internal datastructures for RX */
	mod->rx_num = 0;
	mod->rx_int = 0;

	/* tohost queue descriptors are in page 5 */
	ican3_set_page(mod, QUEUE_TOHOST);
	dst = mod->dpm;

	/* initialize the tohost (rx) queue descriptors: pages 9-24 */
	for (i = 0; i < ICAN3_NEW_BUFFERS; i++) {
		desc.control = DESC_INTERRUPT | DESC_LEN(1); /* I L=1 */
		desc.pointer = mod->free_page;

		/* set wrap flag on last buffer */
		if (i == ICAN3_NEW_BUFFERS - 1)
			desc.control |= DESC_WRAP;

		memcpy_toio(dst, &desc, sizeof(desc));
		dst += sizeof(desc);
		mod->free_page++;
	}

	/* fromhost (tx) mid queue descriptors are in page 6 */
	ican3_set_page(mod, QUEUE_FROMHOST_MID);
	dst = mod->dpm;

	/* setup the internal datastructures for TX */
	mod->tx_num = 0;

	/* initialize the fromhost mid queue descriptors: pages 25-40 */
	for (i = 0; i < ICAN3_NEW_BUFFERS; i++) {
		desc.control = DESC_VALID | DESC_LEN(1); /* V L=1 */
		desc.pointer = mod->free_page;

		/* set wrap flag on last buffer */
		if (i == ICAN3_NEW_BUFFERS - 1)
			desc.control |= DESC_WRAP;

		memcpy_toio(dst, &desc, sizeof(desc));
		dst += sizeof(desc);
		mod->free_page++;
	}

	/* fromhost hi queue descriptors are in page 7 */
	ican3_set_page(mod, QUEUE_FROMHOST_HIGH);
	dst = mod->dpm;

	/* initialize only a single buffer in the fromhost hi queue (unused) */
	desc.control = DESC_VALID | DESC_WRAP | DESC_LEN(1); /* VW L=1 */
	desc.pointer = mod->free_page;
	memcpy_toio(dst, &desc, sizeof(desc));
	mod->free_page++;

	/* fromhost low queue descriptors are in page 8 */
	ican3_set_page(mod, QUEUE_FROMHOST_LOW);
	dst = mod->dpm;

	/* initialize only a single buffer in the fromhost low queue (unused) */
	desc.control = DESC_VALID | DESC_WRAP | DESC_LEN(1); /* VW L=1 */
	desc.pointer = mod->free_page;
	memcpy_toio(dst, &desc, sizeof(desc));
	mod->free_page++;

	spin_unlock_irqrestore(&mod->lock, flags);
}

/*
 * ICAN3 Fast Host Interface Setup
 */

static void ican3_init_fast_host_interface(struct ican3_dev *mod)
{
	struct ican3_fast_desc desc;
	unsigned long flags;
	unsigned int addr;
	void __iomem *dst;
	int i;

	spin_lock_irqsave(&mod->lock, flags);

	/* save the start recv page */
	mod->fastrx_start = mod->free_page;
	mod->fastrx_num = 0;

	/* build a single fast tohost queue descriptor */
	memset(&desc, 0, sizeof(desc));
	desc.control = 0x00;
	desc.command = 1;

	/* build the tohost queue descriptor ring in memory */
	addr = 0;
	for (i = 0; i < ICAN3_RX_BUFFERS; i++) {

		/* set the wrap bit on the last buffer */
		if (i == ICAN3_RX_BUFFERS - 1)
			desc.control |= DESC_WRAP;

		/* switch to the correct page */
		ican3_set_page(mod, mod->free_page);

		/* copy the descriptor to the DPM */
		dst = mod->dpm + addr;
		memcpy_toio(dst, &desc, sizeof(desc));
		addr += sizeof(desc);

		/* move to the next page if necessary */
		if (addr >= DPM_PAGE_SIZE) {
			addr = 0;
			mod->free_page++;
		}
	}

	/* make sure we page-align the next queue */
	if (addr != 0)
		mod->free_page++;

	/* save the start xmit page */
	mod->fasttx_start = mod->free_page;
	mod->fasttx_num = 0;

	/* build a single fast fromhost queue descriptor */
	memset(&desc, 0, sizeof(desc));
	desc.control = DESC_VALID;
	desc.command = 1;

	/* build the fromhost queue descriptor ring in memory */
	addr = 0;
	for (i = 0; i < ICAN3_TX_BUFFERS; i++) {

		/* set the wrap bit on the last buffer */
		if (i == ICAN3_TX_BUFFERS - 1)
			desc.control |= DESC_WRAP;

		/* switch to the correct page */
		ican3_set_page(mod, mod->free_page);

		/* copy the descriptor to the DPM */
		dst = mod->dpm + addr;
		memcpy_toio(dst, &desc, sizeof(desc));
		addr += sizeof(desc);

		/* move to the next page if necessary */
		if (addr >= DPM_PAGE_SIZE) {
			addr = 0;
			mod->free_page++;
		}
	}

	spin_unlock_irqrestore(&mod->lock, flags);
}

/*
 * ICAN3 "new-style" Host Interface Message Helpers
 */

/*
 * LOCKING: must hold mod->lock
 */
static int ican3_new_send_msg(struct ican3_dev *mod, struct ican3_msg *msg)
{
	struct ican3_new_desc desc;
	void __iomem *desc_addr = mod->dpm + (mod->tx_num * sizeof(desc));

	/* switch to the fromhost mid queue, and read the buffer descriptor */
	ican3_set_page(mod, QUEUE_FROMHOST_MID);
	memcpy_fromio(&desc, desc_addr, sizeof(desc));

	if (!(desc.control & DESC_VALID)) {
		netdev_dbg(mod->ndev, "%s: no free buffers\n", __func__);
		return -ENOMEM;
	}

	/* switch to the data page, copy the data */
	ican3_set_page(mod, desc.pointer);
	memcpy_toio(mod->dpm, msg, sizeof(*msg));

	/* switch back to the descriptor, set the valid bit, write it back */
	ican3_set_page(mod, QUEUE_FROMHOST_MID);
	desc.control ^= DESC_VALID;
	memcpy_toio(desc_addr, &desc, sizeof(desc));

	/* update the tx number */
	mod->tx_num = (desc.control & DESC_WRAP) ? 0 : (mod->tx_num + 1);
	return 0;
}

/*
 * LOCKING: must hold mod->lock
 */
static int ican3_new_recv_msg(struct ican3_dev *mod, struct ican3_msg *msg)
{
	struct ican3_new_desc desc;
	void __iomem *desc_addr = mod->dpm + (mod->rx_num * sizeof(desc));

	/* switch to the tohost queue, and read the buffer descriptor */
	ican3_set_page(mod, QUEUE_TOHOST);
	memcpy_fromio(&desc, desc_addr, sizeof(desc));

	if (!(desc.control & DESC_VALID)) {
		netdev_dbg(mod->ndev, "%s: no buffers to recv\n", __func__);
		return -ENOMEM;
	}

	/* switch to the data page, copy the data */
	ican3_set_page(mod, desc.pointer);
	memcpy_fromio(msg, mod->dpm, sizeof(*msg));

	/* switch back to the descriptor, toggle the valid bit, write it back */
	ican3_set_page(mod, QUEUE_TOHOST);
	desc.control ^= DESC_VALID;
	memcpy_toio(desc_addr, &desc, sizeof(desc));

	/* update the rx number */
	mod->rx_num = (desc.control & DESC_WRAP) ? 0 : (mod->rx_num + 1);
	return 0;
}

/*
 * Message Send / Recv Helpers
 */

static int ican3_send_msg(struct ican3_dev *mod, struct ican3_msg *msg)
{
	unsigned long flags;
	int ret;

	spin_lock_irqsave(&mod->lock, flags);

	if (mod->iftype == 0)
		ret = ican3_old_send_msg(mod, msg);
	else
		ret = ican3_new_send_msg(mod, msg);

	spin_unlock_irqrestore(&mod->lock, flags);
	return ret;
}

static int ican3_recv_msg(struct ican3_dev *mod, struct ican3_msg *msg)
{
	unsigned long flags;
	int ret;

	spin_lock_irqsave(&mod->lock, flags);

	if (mod->iftype == 0)
		ret = ican3_old_recv_msg(mod, msg);
	else
		ret = ican3_new_recv_msg(mod, msg);

	spin_unlock_irqrestore(&mod->lock, flags);
	return ret;
}

/*
 * Quick Pre-constructed Messages
 */

static int ican3_msg_connect(struct ican3_dev *mod)
{
	struct ican3_msg msg;

	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_CONNECTI;
	msg.len = cpu_to_le16(0);

	return ican3_send_msg(mod, &msg);
}

static int ican3_msg_disconnect(struct ican3_dev *mod)
{
	struct ican3_msg msg;

	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_DISCONNECT;
	msg.len = cpu_to_le16(0);

	return ican3_send_msg(mod, &msg);
}

static int ican3_msg_newhostif(struct ican3_dev *mod)
{
	struct ican3_msg msg;
	int ret;

	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_NEWHOSTIF;
	msg.len = cpu_to_le16(0);

	/* If we're not using the old interface, switching seems bogus */
	WARN_ON(mod->iftype != 0);

	ret = ican3_send_msg(mod, &msg);
	if (ret)
		return ret;

	/* mark the module as using the new host interface */
	mod->iftype = 1;
	return 0;
}

static int ican3_msg_fasthostif(struct ican3_dev *mod)
{
	struct ican3_msg msg;
	unsigned int addr;

	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_INITFDPMQUEUE;
	msg.len = cpu_to_le16(8);

	/* write the tohost queue start address */
	addr = DPM_PAGE_ADDR(mod->fastrx_start);
	msg.data[0] = addr & 0xff;
	msg.data[1] = (addr >> 8) & 0xff;
	msg.data[2] = (addr >> 16) & 0xff;
	msg.data[3] = (addr >> 24) & 0xff;

	/* write the fromhost queue start address */
	addr = DPM_PAGE_ADDR(mod->fasttx_start);
	msg.data[4] = addr & 0xff;
	msg.data[5] = (addr >> 8) & 0xff;
	msg.data[6] = (addr >> 16) & 0xff;
	msg.data[7] = (addr >> 24) & 0xff;

	/* If we're not using the new interface yet, we cannot do this */
	WARN_ON(mod->iftype != 1);

	return ican3_send_msg(mod, &msg);
}

/*
 * Setup the CAN filter to either accept or reject all
 * messages from the CAN bus.
 */
static int ican3_set_id_filter(struct ican3_dev *mod, bool accept)
{
	struct ican3_msg msg;
	int ret;

	/* Standard Frame Format */
	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_SETAFILMASK;
	msg.len = cpu_to_le16(5);
	msg.data[0] = 0x00; /* IDLo LSB */
	msg.data[1] = 0x00; /* IDLo MSB */
	msg.data[2] = 0xff; /* IDHi LSB */
	msg.data[3] = 0x07; /* IDHi MSB */

	/* accept all frames for fast host if, or reject all frames */
	msg.data[4] = accept ? SETAFILMASK_FASTIF : SETAFILMASK_REJECT;

	ret = ican3_send_msg(mod, &msg);
	if (ret)
		return ret;

	/* Extended Frame Format */
	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_SETAFILMASK;
	msg.len = cpu_to_le16(13);
	msg.data[0] = 0;    /* MUX = 0 */
	msg.data[1] = 0x00; /* IDLo LSB */
	msg.data[2] = 0x00;
	msg.data[3] = 0x00;
	msg.data[4] = 0x20; /* IDLo MSB */
	msg.data[5] = 0xff; /* IDHi LSB */
	msg.data[6] = 0xff;
	msg.data[7] = 0xff;
	msg.data[8] = 0x3f; /* IDHi MSB */

	/* accept all frames for fast host if, or reject all frames */
	msg.data[9] = accept ? SETAFILMASK_FASTIF : SETAFILMASK_REJECT;

	return ican3_send_msg(mod, &msg);
}

/*
 * Bring the CAN bus online or offline
 */
static int ican3_set_bus_state(struct ican3_dev *mod, bool on)
{
	struct can_bittiming *bt = &mod->can.bittiming;
	struct ican3_msg msg;
	u8 btr0, btr1;
	int res;

	/* This algorithm was stolen from drivers/net/can/sja1000/sja1000.c      */
	/* The bittiming register command for the ICAN3 just sets the bit timing */
	/* registers on the SJA1000 chip directly                                */
	btr0 = ((bt->brp - 1) & 0x3f) | (((bt->sjw - 1) & 0x3) << 6);
	btr1 = ((bt->prop_seg + bt->phase_seg1 - 1) & 0xf) |
		(((bt->phase_seg2 - 1) & 0x7) << 4);
	if (mod->can.ctrlmode & CAN_CTRLMODE_3_SAMPLES)
		btr1 |= 0x80;

	if (mod->fwtype == ICAN3_FWTYPE_ICANOS) {
		if (on) {
			/* set bittiming */
			memset(&msg, 0, sizeof(msg));
			msg.spec = MSG_CBTRREQ;
			msg.len = cpu_to_le16(4);
			msg.data[0] = 0x00;
			msg.data[1] = 0x00;
			msg.data[2] = btr0;
			msg.data[3] = btr1;

			res = ican3_send_msg(mod, &msg);
			if (res)
				return res;
		}

		/* can-on/off request */
		memset(&msg, 0, sizeof(msg));
		msg.spec = on ? MSG_CONREQ : MSG_COFFREQ;
		msg.len = cpu_to_le16(0);

		return ican3_send_msg(mod, &msg);

	} else if (mod->fwtype == ICAN3_FWTYPE_CAL_CANOPEN) {
		/* bittiming + can-on/off request */
		memset(&msg, 0, sizeof(msg));
		msg.spec = MSG_LMTS;
		if (on) {
			msg.len = cpu_to_le16(4);
			msg.data[0] = LMTS_BUSON_REQ;
			msg.data[1] = 0;
			msg.data[2] = btr0;
			msg.data[3] = btr1;
		} else {
			msg.len = cpu_to_le16(2);
			msg.data[0] = LMTS_BUSOFF_REQ;
			msg.data[1] = 0;
		}
		res = ican3_send_msg(mod, &msg);
		if (res)
			return res;

		if (on) {
			/* create NMT Slave Node for error processing
			 *   class 2 (with error capability, see CiA/DS203-1)
			 *   id    1
			 *   name  locnod1 (must be exactly 7 bytes)
			 */
			memset(&msg, 0, sizeof(msg));
			msg.spec = MSG_NMTS;
			msg.len = cpu_to_le16(11);
			msg.data[0] = NMTS_CREATE_NODE_REQ;
			msg.data[1] = 0;
			msg.data[2] = 2;                 /* node class */
			msg.data[3] = 1;                 /* node id */
			strcpy(msg.data + 4, "locnod1"); /* node name  */
			return ican3_send_msg(mod, &msg);
		}
		return 0;
	}
	return -ENOTSUPP;
}

static int ican3_set_termination(struct ican3_dev *mod, bool on)
{
	struct ican3_msg msg;

	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_HWCONF;
	msg.len = cpu_to_le16(2);
	msg.data[0] = 0x00;
	msg.data[1] = on ? HWCONF_TERMINATE_ON : HWCONF_TERMINATE_OFF;

	return ican3_send_msg(mod, &msg);
}

static int ican3_send_inquiry(struct ican3_dev *mod, u8 subspec)
{
	struct ican3_msg msg;

	memset(&msg, 0, sizeof(msg));
	msg.spec = MSG_INQUIRY;
	msg.len = cpu_to_le16(2);
	msg.data[0] = subspec;
	msg.data[1] = 0x00;

	return ican3_send_msg(mod, &msg);
}

static int ican3_set_buserror(struct ican3_dev *mod, u8 quota)
{
	struct ican3_msg msg;

	if (mod->fwtype == ICAN3_FWTYPE_ICANOS) {
		memset(&msg, 0, sizeof(msg));
		msg.spec = MSG_CCONFREQ;
		msg.len = cpu_to_le16(2);
		msg.data[0] = 0x00;
		msg.data[1] = quota;
	} else if (mod->fwtype == ICAN3_FWTYPE_CAL_CANOPEN) {
		memset(&msg, 0, sizeof(msg));
		msg.spec = MSG_LMTS;
		msg.len = cpu_to_le16(4);
		msg.data[0] = LMTS_CAN_CONF_REQ;
		msg.data[1] = 0x00;
		msg.data[2] = 0x00;
		msg.data[3] = quota;
	} else {
		return -ENOTSUPP;
	}
	return ican3_send_msg(mod, &msg);
}

/*
 * ICAN3 to Linux CAN Frame Conversion
 */

static void ican3_to_can_frame(struct ican3_dev *mod,
			       struct ican3_fast_desc *desc,
			       struct can_frame *cf)
{
	if ((desc->command & ICAN3_CAN_TYPE_MASK) == ICAN3_CAN_TYPE_SFF) {
		if (desc->data[1] & ICAN3_SFF_RTR)
			cf->can_id |= CAN_RTR_FLAG;

		cf->can_id |= desc->data[0] << 3;
		cf->can_id |= (desc->data[1] & 0xe0) >> 5;
		cf->can_dlc = get_can_dlc(desc->data[1] & ICAN3_CAN_DLC_MASK);