mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	stable@vger.kernel.org, Ilya Dryomov <idryomov@gmail.com>,
	Josh Durgin <jdurgin@redhat.com>
Subject: [PATCH 4.4 64/68] libceph: apply new_state before new_up_client on incrementals
Date: Mon,  8 Aug 2016 21:11:39 +0200	[thread overview]
Message-ID: <20160808180212.958987261@linuxfoundation.org> (raw)
In-Reply-To: <20160808180209.697765393@linuxfoundation.org>

4.4-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Ilya Dryomov <idryomov@gmail.com>

commit 930c532869774ebf8af9efe9484c597f896a7d46 upstream.

Currently, osd_weight and osd_state fields are updated in the encoding
order.  This is wrong, because an incremental map may look like e.g.

    new_up_client: { osd=6, addr=... } # set osd_state and addr
    new_state: { osd=6, xorstate=EXISTS } # clear osd_state

Suppose osd6's current osd_state is EXISTS (i.e. osd6 is down).  After
applying new_up_client, osd_state is changed to EXISTS | UP.  Carrying
on with the new_state update, we flip EXISTS and leave osd6 in a weird
"!EXISTS but UP" state.  A non-existent OSD is considered down by the
mapping code

2087    for (i = 0; i < pg->pg_temp.len; i++) {
2088            if (ceph_osd_is_down(osdmap, pg->pg_temp.osds[i])) {
2089                    if (ceph_can_shift_osds(pi))
2090                            continue;
2091
2092                    temp->osds[temp->size++] = CRUSH_ITEM_NONE;

and so requests get directed to the second OSD in the set instead of
the first, resulting in OSD-side errors like:

[WRN] : client.4239 192.168.122.21:0/2444980242 misdirected client.4239.1:2827 pg 2.5df899f2 to osd.4 not [1,4,6] in e680/680

and hung rbds on the client:

[  493.566367] rbd: rbd0: write 400000 at 11cc00000 (0)
[  493.566805] rbd: rbd0:   result -6 xferred 400000
[  493.567011] blk_update_request: I/O error, dev rbd0, sector 9330688

The fix is to decouple application from the decoding and:
- apply new_weight first
- apply new_state before new_up_client
- twiddle osd_state flags if marking in
- clear out some of the state if osd is destroyed

Fixes: http://tracker.ceph.com/issues/14901

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

---
 net/ceph/osdmap.c |  156 +++++++++++++++++++++++++++++++++++++++---------------
 1 file changed, 113 insertions(+), 43 deletions(-)

--- a/net/ceph/osdmap.c
+++ b/net/ceph/osdmap.c
@@ -1192,6 +1192,115 @@ struct ceph_osdmap *ceph_osdmap_decode(v
 }
 
 /*
+ * Encoding order is (new_up_client, new_state, new_weight).  Need to
+ * apply in the (new_weight, new_state, new_up_client) order, because
+ * an incremental map may look like e.g.
+ *
+ *     new_up_client: { osd=6, addr=... } # set osd_state and addr
+ *     new_state: { osd=6, xorstate=EXISTS } # clear osd_state
+ */
+static int decode_new_up_state_weight(void **p, void *end,
+				      struct ceph_osdmap *map)
+{
+	void *new_up_client;
+	void *new_state;
+	void *new_weight_end;
+	u32 len;
+
+	new_up_client = *p;
+	ceph_decode_32_safe(p, end, len, e_inval);
+	len *= sizeof(u32) + sizeof(struct ceph_entity_addr);
+	ceph_decode_need(p, end, len, e_inval);
+	*p += len;
+
+	new_state = *p;
+	ceph_decode_32_safe(p, end, len, e_inval);
+	len *= sizeof(u32) + sizeof(u8);
+	ceph_decode_need(p, end, len, e_inval);
+	*p += len;
+
+	/* new_weight */
+	ceph_decode_32_safe(p, end, len, e_inval);
+	while (len--) {
+		s32 osd;
+		u32 w;
+
+		ceph_decode_need(p, end, 2*sizeof(u32), e_inval);
+		osd = ceph_decode_32(p);
+		w = ceph_decode_32(p);
+		BUG_ON(osd >= map->max_osd);
+		pr_info("osd%d weight 0x%x %s\n", osd, w,
+		     w == CEPH_OSD_IN ? "(in)" :
+		     (w == CEPH_OSD_OUT ? "(out)" : ""));
+		map->osd_weight[osd] = w;
+
+		/*
+		 * If we are marking in, set the EXISTS, and clear the
+		 * AUTOOUT and NEW bits.
+		 */
+		if (w) {
+			map->osd_state[osd] |= CEPH_OSD_EXISTS;
+			map->osd_state[osd] &= ~(CEPH_OSD_AUTOOUT |
+						 CEPH_OSD_NEW);
+		}
+	}
+	new_weight_end = *p;
+
+	/* new_state (up/down) */
+	*p = new_state;
+	len = ceph_decode_32(p);
+	while (len--) {
+		s32 osd;
+		u8 xorstate;
+		int ret;
+
+		osd = ceph_decode_32(p);
+		xorstate = ceph_decode_8(p);
+		if (xorstate == 0)
+			xorstate = CEPH_OSD_UP;
+		BUG_ON(osd >= map->max_osd);
+		if ((map->osd_state[osd] & CEPH_OSD_UP) &&
+		    (xorstate & CEPH_OSD_UP))
+			pr_info("osd%d down\n", osd);
+		if ((map->osd_state[osd] & CEPH_OSD_EXISTS) &&
+		    (xorstate & CEPH_OSD_EXISTS)) {
+			pr_info("osd%d does not exist\n", osd);
+			map->osd_weight[osd] = CEPH_OSD_IN;
+			ret = set_primary_affinity(map, osd,
+						   CEPH_OSD_DEFAULT_PRIMARY_AFFINITY);
+			if (ret)
+				return ret;
+			memset(map->osd_addr + osd, 0, sizeof(*map->osd_addr));
+			map->osd_state[osd] = 0;
+		} else {
+			map->osd_state[osd] ^= xorstate;
+		}
+	}
+
+	/* new_up_client */
+	*p = new_up_client;
+	len = ceph_decode_32(p);
+	while (len--) {
+		s32 osd;
+		struct ceph_entity_addr addr;
+
+		osd = ceph_decode_32(p);
+		ceph_decode_copy(p, &addr, sizeof(addr));
+		ceph_decode_addr(&addr);
+		BUG_ON(osd >= map->max_osd);
+		pr_info("osd%d up\n", osd);
+		map->osd_state[osd] |= CEPH_OSD_EXISTS | CEPH_OSD_UP;
+		map->osd_addr[osd] = addr;
+	}
+
+	*p = new_weight_end;
+	return 0;
+
+e_inval:
+	return -EINVAL;
+}
+
+/*
  * decode and apply an incremental map update.
  */
 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
@@ -1290,49 +1399,10 @@ struct ceph_osdmap *osdmap_apply_increme
 			__remove_pg_pool(&map->pg_pools, pi);
 	}
 
-	/* new_up */
-	ceph_decode_32_safe(p, end, len, e_inval);
-	while (len--) {
-		u32 osd;
-		struct ceph_entity_addr addr;
-		ceph_decode_32_safe(p, end, osd, e_inval);
-		ceph_decode_copy_safe(p, end, &addr, sizeof(addr), e_inval);
-		ceph_decode_addr(&addr);
-		pr_info("osd%d up\n", osd);
-		BUG_ON(osd >= map->max_osd);
-		map->osd_state[osd] |= CEPH_OSD_UP | CEPH_OSD_EXISTS;
-		map->osd_addr[osd] = addr;
-	}
-
-	/* new_state */
-	ceph_decode_32_safe(p, end, len, e_inval);
-	while (len--) {
-		u32 osd;
-		u8 xorstate;
-		ceph_decode_32_safe(p, end, osd, e_inval);
-		xorstate = **(u8 **)p;
-		(*p)++;  /* clean flag */
-		if (xorstate == 0)
-			xorstate = CEPH_OSD_UP;
-		if (xorstate & CEPH_OSD_UP)
-			pr_info("osd%d down\n", osd);
-		if (osd < map->max_osd)
-			map->osd_state[osd] ^= xorstate;
-	}
-
-	/* new_weight */
-	ceph_decode_32_safe(p, end, len, e_inval);
-	while (len--) {
-		u32 osd, off;
-		ceph_decode_need(p, end, sizeof(u32)*2, e_inval);
-		osd = ceph_decode_32(p);
-		off = ceph_decode_32(p);
-		pr_info("osd%d weight 0x%x %s\n", osd, off,
-		     off == CEPH_OSD_IN ? "(in)" :
-		     (off == CEPH_OSD_OUT ? "(out)" : ""));
-		if (osd < map->max_osd)
-			map->osd_weight[osd] = off;
-	}
+	/* new_up_client, new_state, new_weight */
+	err = decode_new_up_state_weight(p, end, map);
+	if (err)
+		goto bad;
 
 	/* new_pg_temp */
 	err = decode_new_pg_temp(p, end, map);

  parent reply	other threads:[~2016-08-08 19:45 UTC|newest]

Thread overview: 68+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <CGME20160808191310uscas1p28a09c69fbfd2c15b26ac1015e4a2d81c@uscas1p2.samsung.com>
2016-08-08 19:10 ` [PATCH 4.4 00/68] 4.4.17-stable review Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 02/68] x86/quirks: Apply nvidia_bugs quirk only on root bus Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 03/68] x86/quirks: Reintroduce scanning of secondary buses Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 05/68] dmaengine: at_xdmac: align descriptors on 64 bits Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 06/68] dmaengine: at_xdmac: fix residue corruption Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 07/68] dmaengine: at_xdmac: double FIFO flush needed to compute residue Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 08/68] mm, sl[au]b: add __GFP_ATOMIC to the GFP reclaim mask Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 09/68] mm, compaction: abort free scanner if split fails Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 10/68] fs/nilfs2: fix potential underflow in call to crc32_le Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 11/68] mm, compaction: prevent VM_BUG_ON when terminating freeing scanner Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 12/68] mm, meminit: always return a valid node from early_pfn_to_nid Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 13/68] mm, meminit: ensure node is online before checking whether pages are uninitialised Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 14/68] vmlinux.lds: account for destructor sections Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 15/68] pps: do not crash when failed to register Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 16/68] kernel/sysrq, watchdog, sched/core: Reset watchdog on all CPUs while processing sysrq-w Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 17/68] arc: unwind: warn only once if DW2_UNWIND is disabled Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 18/68] ARC: unwind: ensure that .debug_frame is generated (vs. .eh_frame) Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 19/68] xen/pciback: Fix conf_space read/write overlap check Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 20/68] xenbus: dont BUG() on user mode induced condition Greg Kroah-Hartman
2016-08-08 19:10   ` [PATCH 4.4 21/68] xenbus: dont bail early from xenbus_dev_request_and_reply() Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 25/68] Input: vmmouse - remove port reservation Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 26/68] Input: elantech - add more IC body types to the list Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 27/68] Input: xpad - fix oops when attaching an unknown Xbox One gamepad Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 28/68] Input: wacom_w8001 - w8001_MAX_LENGTH should be 13 Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 29/68] Input: xpad - validate USB endpoint count during probe Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 31/68] pvclock: Add CPU barriers to get correct version value Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 32/68] pinctrl: single: Fix missing flush of posted write for a wakeirq Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 33/68] pinctrl: imx: Do not treat a PIN without MUX register as an error Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 34/68] cgroup: set css->id to -1 during init Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 35/68] power_supply: power_supply_read_temp only if use_cnt > 0 Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 36/68] locks: use file_inode() Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 37/68] Revert "ecryptfs: forbid opening files without mmap handler" Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 38/68] ecryptfs: dont allow mmap when the lower fs doesnt support it Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 39/68] ext4: verify extent header depth Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 40/68] 9p: use file_dentry() Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 41/68] namespace: update event counter when umounting a deleted dentry Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 42/68] spi: sunxi: fix transfer timeout Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 43/68] spi: sun4i: fix FIFO limit Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 44/68] clk: rockchip: initialize flags of clk_init_data in mmc-phase clock Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 45/68] platform/chrome: cros_ec_dev - double fetch bug in ioctl Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 46/68] qeth: delete napi struct when removing a qeth device Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 47/68] block: fix use-after-free in sys_ioprio_get() Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 48/68] mmc: block: fix packed command header endianness Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 49/68] sched/fair: Fix effective_load() to consistently use smoothed load Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 50/68] ovl: handle ATTR_KILL* Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 51/68] perf/x86: fix PEBS issues on Intel Atom/Core2 Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 52/68] can: at91_can: RX queue could get stuck at high bus load Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 53/68] can: c_can: Update D_CAN TX and RX functions to 32 bit - fix Altera Cyclone access Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 54/68] can: fix handling of unmodifiable configuration options fix Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 55/68] can: fix oops caused by wrong rtnl dellink usage Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 56/68] RDS: fix rds_tcp_init() error path Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 57/68] SCSI: fix new bug in scsi_dev_info_list string matching Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 58/68] ipr: Clear interrupt on croc/crocodile when running with LSI Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 59/68] media: fix airspy usb probe error path Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 60/68] posix_cpu_timer: Exit early when process has been reaped Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 61/68] i2c: qup: Fix wrong value of index variable Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 62/68] i2c: mux: reg: wrong condition checked for of_address_to_resource return value Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 63/68] libata: LITE-ON CX1-JB256-HP needs lower max_sectors Greg Kroah-Hartman
2016-08-08 19:11   ` Greg Kroah-Hartman [this message]
2016-08-08 19:11   ` [PATCH 4.4 65/68] net: mvneta: set real interrupt per packet for tx_done Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 66/68] intel_th: pci: Add Kaby Lake PCH-H support Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 67/68] intel_th: Fix a deadlock in modprobing Greg Kroah-Hartman
2016-08-08 19:11   ` [PATCH 4.4 68/68] vfs: fix deadlock in file_remove_privs() on overlayfs Greg Kroah-Hartman
2016-08-09  4:22   ` [PATCH 4.4 00/68] 4.4.17-stable review Guenter Roeck
2016-08-09  8:21     ` Greg Kroah-Hartman
2016-08-09 16:14       ` Guenter Roeck
2016-08-09 17:22         ` Greg Kroah-Hartman
2016-08-09 15:11   ` Shuah Khan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160808180212.958987261@linuxfoundation.org \
    --to=gregkh@linuxfoundation.org \
    --cc=idryomov@gmail.com \
    --cc=jdurgin@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

Powered by JetHome