From patchwork Thu Apr 13 21:25:17 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Dave Jiang X-Patchwork-Id: 9680035 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id C9FBA60381 for ; Thu, 13 Apr 2017 21:25:23 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id B67FE28664 for ; Thu, 13 Apr 2017 21:25:23 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id A7BDF286AC; Thu, 13 Apr 2017 21:25:23 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-1.9 required=2.0 tests=BAYES_00, RCVD_IN_DNSWL_NONE autolearn=ham version=3.3.1 Received: from ml01.01.org (ml01.01.org [198.145.21.10]) (using TLSv1.2 with cipher DHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.wl.linuxfoundation.org (Postfix) with ESMTPS id EE97D28664 for ; Thu, 13 Apr 2017 21:25:22 +0000 (UTC) Received: from [127.0.0.1] (localhost [IPv6:::1]) by ml01.01.org (Postfix) with ESMTP id 9881C2194EB5C; Thu, 13 Apr 2017 14:25:22 -0700 (PDT) X-Original-To: linux-nvdimm@lists.01.org Delivered-To: linux-nvdimm@lists.01.org Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ml01.01.org (Postfix) with ESMTPS id 6A46A21A04830 for ; Thu, 13 Apr 2017 14:25:20 -0700 (PDT) Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by fmsmga105.fm.intel.com with ESMTP; 13 Apr 2017 14:25:17 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.37,195,1488873600"; d="scan'208";a="845606617" Received: from djiang5-desk3.ch.intel.com ([143.182.137.38]) by FMSMGA003.fm.intel.com with ESMTP; 13 Apr 2017 14:25:17 -0700 Subject: [PATCH v6] libnvdimm: fix clear poison locking with spinlock and GFP_NOWAIT allocation From: Dave Jiang To: dan.j.williams@intel.com Date: Thu, 13 Apr 2017 14:25:17 -0700 Message-ID: <149211841881.15608.18107264434663008282.stgit@djiang5-desk3.ch.intel.com> User-Agent: StGit/0.17.1-dirty MIME-Version: 1.0 X-BeenThere: linux-nvdimm@lists.01.org X-Mailman-Version: 2.1.22 Precedence: list List-Id: "Linux-nvdimm developer list." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: linux-nvdimm@lists.01.org Errors-To: linux-nvdimm-bounces@lists.01.org Sender: "Linux-nvdimm" X-Virus-Scanned: ClamAV using ClamSMTP The following warning results from holding a lane spinlock, preempt_disable(), or the btt map spinlock and then trying to take the reconfig_mutex to walk the poison list and potentially add new entries. BUG: sleeping function called from invalid context at kernel/locking/mutex. c:747 in_atomic(): 1, irqs_disabled(): 0, pid: 17159, name: dd [..] Call Trace: dump_stack+0x85/0xc8 ___might_sleep+0x184/0x250 __might_sleep+0x4a/0x90 __mutex_lock+0x58/0x9b0 ? nvdimm_bus_lock+0x21/0x30 [libnvdimm] ? __nvdimm_bus_badblocks_clear+0x2f/0x60 [libnvdimm] ? acpi_nfit_forget_poison+0x79/0x80 [nfit] ? _raw_spin_unlock+0x27/0x40 mutex_lock_nested+0x1b/0x20 nvdimm_bus_lock+0x21/0x30 [libnvdimm] nvdimm_forget_poison+0x25/0x50 [libnvdimm] nvdimm_clear_poison+0x106/0x140 [libnvdimm] nsio_rw_bytes+0x164/0x270 [libnvdimm] btt_write_pg+0x1de/0x3e0 [nd_btt] ? blk_queue_enter+0x30/0x290 btt_make_request+0x11a/0x310 [nd_btt] ? blk_queue_enter+0xb7/0x290 ? blk_queue_enter+0x30/0x290 generic_make_request+0x118/0x3b0 A spinlock is introduced to protect the poison list. This allows us to not having to acquire the reconfig_mutex for touching the poison list. The add_poison() function has been broken out into two helper functions. One to allocate the poison entry and the other to apppend the entry. This allows us to unlock the poison_lock in non-I/O path and continue to be able to allocate the poison entry with GFP_KERNEL. We will use GFP_NOWAIT in the I/O path in order to satisfy being in atomic context. Signed-off-by: Dave Jiang --- v2: Dan - remove mempool and just use GFP_NOWAIT v3: Vishal - remove unneeded code v4: Dan - fixup commit header - GFP_NOWAIT only in I/O path v5: Dan - address issue where when lock released to allocate, the list has updated v6: Dan - open code poison entry allocation - remove NULL check for kfree() drivers/nvdimm/bus.c | 7 +++--- drivers/nvdimm/core.c | 56 +++++++++++++++++++++++++++------------------ drivers/nvdimm/nd-core.h | 1 + include/linux/libnvdimm.h | 2 -- 4 files changed, 38 insertions(+), 28 deletions(-) diff --git a/drivers/nvdimm/bus.c b/drivers/nvdimm/bus.c index 5ad2e59..d214ac44 100644 --- a/drivers/nvdimm/bus.c +++ b/drivers/nvdimm/bus.c @@ -296,6 +296,7 @@ struct nvdimm_bus *nvdimm_bus_register(struct device *parent, init_waitqueue_head(&nvdimm_bus->probe_wait); nvdimm_bus->id = ida_simple_get(&nd_ida, 0, 0, GFP_KERNEL); mutex_init(&nvdimm_bus->reconfig_mutex); + spin_lock_init(&nvdimm_bus->poison_lock); if (nvdimm_bus->id < 0) { kfree(nvdimm_bus); return NULL; @@ -364,9 +365,9 @@ static int nd_bus_remove(struct device *dev) nd_synchronize(); device_for_each_child(&nvdimm_bus->dev, NULL, child_unregister); - nvdimm_bus_lock(&nvdimm_bus->dev); + spin_lock(&nvdimm_bus->poison_lock); free_poison_list(&nvdimm_bus->poison_list); - nvdimm_bus_unlock(&nvdimm_bus->dev); + spin_unlock(&nvdimm_bus->poison_lock); nvdimm_bus_destroy_ndctl(nvdimm_bus); @@ -990,7 +991,7 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm, if (clear_err->cleared) { /* clearing the poison list we keep track of */ - __nvdimm_forget_poison(nvdimm_bus, clear_err->address, + nvdimm_forget_poison(nvdimm_bus, clear_err->address, clear_err->cleared); /* now sync the badblocks lists */ diff --git a/drivers/nvdimm/core.c b/drivers/nvdimm/core.c index 40a3da0..2dee908 100644 --- a/drivers/nvdimm/core.c +++ b/drivers/nvdimm/core.c @@ -518,6 +518,15 @@ void nvdimm_badblocks_populate(struct nd_region *nd_region, } EXPORT_SYMBOL_GPL(nvdimm_badblocks_populate); +static void append_poison_entry(struct nvdimm_bus *nvdimm_bus, + struct nd_poison *pl, u64 addr, u64 length) +{ + lockdep_assert_held(&nvdimm_bus->poison_lock); + pl->start = addr; + pl->length = length; + list_add_tail(&pl->list, &nvdimm_bus->poison_list); +} + static int add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length, gfp_t flags) { @@ -527,19 +536,24 @@ static int add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length, if (!pl) return -ENOMEM; - pl->start = addr; - pl->length = length; - list_add_tail(&pl->list, &nvdimm_bus->poison_list); - + append_poison_entry(nvdimm_bus, pl, addr, length); return 0; } static int bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length) { - struct nd_poison *pl; + struct nd_poison *pl, *pl_new; - if (list_empty(&nvdimm_bus->poison_list)) - return add_poison(nvdimm_bus, addr, length, GFP_KERNEL); + spin_unlock(&nvdimm_bus->poison_lock); + pl_new = kzalloc(sizeof(*pl_new), GFP_KERNEL); + spin_lock(&nvdimm_bus->poison_lock); + + if (list_empty(&nvdimm_bus->poison_list)) { + if (!pl_new) + return -ENOMEM; + append_poison_entry(nvdimm_bus, pl_new, addr, length); + return 0; + } /* * There is a chance this is a duplicate, check for those first. @@ -551,6 +565,7 @@ static int bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length) /* If length has changed, update this list entry */ if (pl->length != length) pl->length = length; + kfree(pl_new); return 0; } @@ -559,30 +574,33 @@ static int bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length) * as any overlapping ranges will get resolved when the list is consumed * and converted to badblocks */ - return add_poison(nvdimm_bus, addr, length, GFP_KERNEL); + if (!pl_new) + return -ENOMEM; + append_poison_entry(nvdimm_bus, pl_new, addr, length); + + return 0; } int nvdimm_bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length) { int rc; - nvdimm_bus_lock(&nvdimm_bus->dev); + spin_lock(&nvdimm_bus->poison_lock); rc = bus_add_poison(nvdimm_bus, addr, length); - nvdimm_bus_unlock(&nvdimm_bus->dev); + spin_unlock(&nvdimm_bus->poison_lock); return rc; } EXPORT_SYMBOL_GPL(nvdimm_bus_add_poison); -void __nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, phys_addr_t start, +void nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, phys_addr_t start, unsigned int len) { struct list_head *poison_list = &nvdimm_bus->poison_list; u64 clr_end = start + len - 1; struct nd_poison *pl, *next; - lockdep_assert_held(&nvdimm_bus->reconfig_mutex); - + spin_lock(&nvdimm_bus->poison_lock); WARN_ON_ONCE(list_empty(poison_list)); /* @@ -629,21 +647,13 @@ void __nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, phys_addr_t start, u64 new_len = pl_end - new_start + 1; /* Add new entry covering the right half */ - add_poison(nvdimm_bus, new_start, new_len, GFP_NOIO); + add_poison(nvdimm_bus, new_start, new_len, GFP_NOWAIT); /* Adjust this entry to cover the left half */ pl->length = start - pl->start; continue; } } -} -EXPORT_SYMBOL_GPL(__nvdimm_forget_poison); - -void nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, - phys_addr_t start, unsigned int len) -{ - nvdimm_bus_lock(&nvdimm_bus->dev); - __nvdimm_forget_poison(nvdimm_bus, start, len); - nvdimm_bus_unlock(&nvdimm_bus->dev); + spin_unlock(&nvdimm_bus->poison_lock); } EXPORT_SYMBOL_GPL(nvdimm_forget_poison); diff --git a/drivers/nvdimm/nd-core.h b/drivers/nvdimm/nd-core.h index 8623e57..4c4bd20 100644 --- a/drivers/nvdimm/nd-core.h +++ b/drivers/nvdimm/nd-core.h @@ -32,6 +32,7 @@ struct nvdimm_bus { struct list_head poison_list; struct list_head mapping_list; struct mutex reconfig_mutex; + spinlock_t poison_lock; }; struct nvdimm { diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h index 1c609e8..98b2076 100644 --- a/include/linux/libnvdimm.h +++ b/include/linux/libnvdimm.h @@ -122,8 +122,6 @@ static inline struct nd_blk_region_desc *to_blk_region_desc( int nvdimm_bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length); void nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, phys_addr_t start, unsigned int len); -void __nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, - phys_addr_t start, unsigned int len); struct nvdimm_bus *nvdimm_bus_register(struct device *parent, struct nvdimm_bus_descriptor *nfit_desc); void nvdimm_bus_unregister(struct nvdimm_bus *nvdimm_bus);