Oleg reports a division by zero error on zero-length write() to the percpu_pagelist_fraction sysctl: divide error: 0000 [#1] SMP DEBUG_PAGEALLOC CPU: 1 PID: 9142 Comm: badarea_io Not tainted 3.15.0-rc2-vm-nfs+ #19 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011 task: ffff8800d5aeb6e0 ti: ffff8800d87a2000 task.ti: ffff8800d87a2000 RIP: 0010: percpu_pagelist_fraction_sysctl_handler+0x84/0x120 RSP: 0018:ffff8800d87a3e78 EFLAGS: 00010246 RAX: 0000000000000f89 RBX: ffff88011f7fd000 RCX: 0000000000000000 RDX: 0000000000000000 RSI: 0000000000000001 RDI: 0000000000000010 RBP: ffff8800d87a3e98 R08: ffffffff81d002c8 R09: ffff8800d87a3f50 R10: 000000000000000b R11: 0000000000000246 R12: 0000000000000060 R13: ffffffff81c3c3e0 R14: ffffffff81cfddf8 R15: ffff8801193b0800 FS: 00007f614f1e9740(0000) GS:ffff88011f440000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f614f1fa000 CR3: 00000000d9291000 CR4: 00000000000006e0 Call Trace: proc_sys_call_handler+0xb3/0xc0 proc_sys_write+0x14/0x20 vfs_write+0xba/0x1e0 SyS_write+0x46/0xb0 tracesys+0xe1/0xe6 However, if the percpu_pagelist_fraction sysctl is set by the user, it is also impossible to restore it to the kernel default since the user cannot write 0 to the sysctl. This patch allows the user to write 0 to restore the default behavior. It still requires a fraction equal to or larger than 8, however, as stated by the documentation for sanity. If a value in the range [1, 7] is written, the sysctl will return EINVAL. This successfully solves the divide by zero issue at the same time.
Unified · Split
| @@ -702,7 +702,8 @@ The batch value of each per cpu pagelist is also updated as a result. It is · 701 unmodified lines | ||
| 702 | 702 | set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) |
| 703 | 703 | |
| 704 | 704 | The initial value is zero. Kernel does not use this value at boot time to set |
| 705 | −the high water marks for each per cpu page list. | |
| 705 | +the high water marks for each per cpu page list. If the user writes '0' to this | |
| 706 | +sysctl, it will revert to this default behavior. | |
| 706 | 707 | |
| 707 | 708 | ============================================================== |
| 708 | 709 | |
| @@ -136,7 +136,6 @@ static unsigned long dirty_bytes_min = 2 * PAGE_SIZE; · 135 unmodified lines | ||
| 136 | 136 | /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */ |
| 137 | 137 | static int maxolduid = 65535; |
| 138 | 138 | static int minolduid; |
| 139 | −static int min_percpu_pagelist_fract = 8; | |
| 140 | 139 | |
| 141 | 140 | static int ngroups_max = NGROUPS_MAX; |
| 142 | 141 | static const int cap_last_cap = CAP_LAST_CAP; |
| @@ -1317,7 +1316,7 @@ static struct ctl_table vm_table[] = { · 1174 unmodified lines | ||
| 1317 | 1316 | .maxlen = sizeof(percpu_pagelist_fraction), |
| 1318 | 1317 | .mode = 0644, |
| 1319 | 1318 | .proc_handler = percpu_pagelist_fraction_sysctl_handler, |
| 1320 | − .extra1 = &min_percpu_pagelist_fract, | |
| 1319 | + .extra1 = &zero, | |
| 1321 | 1320 | }, |
| 1322 | 1321 | #ifdef CONFIG_MMU |
| 1323 | 1322 | { |
| @@ -69,6 +69,7 @@ · 68 unmodified lines | ||
| 69 | 69 | |
| 70 | 70 | /* prevent >1 _updater_ of zone percpu pageset ->high and ->batch fields */ |
| 71 | 71 | static DEFINE_MUTEX(pcp_batch_high_lock); |
| 72 | +#define MIN_PERCPU_PAGELIST_FRACTION (8) | |
| 72 | 73 | |
| 73 | 74 | #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID |
| 74 | 75 | DEFINE_PER_CPU(int, numa_node); |
| @@ -4145,7 +4146,7 @@ static void __meminit zone_init_free_lists(struct zone *zone) · 4070 unmodified lines | ||
| 4145 | 4146 | memmap_init_zone((size), (nid), (zone), (start_pfn), MEMMAP_EARLY) |
| 4146 | 4147 | #endif |
| 4147 | 4148 | |
| 4148 | −static int __meminit zone_batchsize(struct zone *zone) | |
| 4149 | +static int zone_batchsize(struct zone *zone) | |
| 4149 | 4150 | { |
| 4150 | 4151 | #ifdef CONFIG_MMU |
| 4151 | 4152 | int batch; |
| @@ -4261,8 +4262,8 @@ static void pageset_set_high(struct per_cpu_pageset *p, · 109 unmodified lines | ||
| 4261 | 4262 | pageset_update(&p->pcp, high, batch); |
| 4262 | 4263 | } |
| 4263 | 4264 | |
| 4264 | −static void __meminit pageset_set_high_and_batch(struct zone *zone, | |
| 4265 | − struct per_cpu_pageset *pcp) | |
| 4265 | +static void pageset_set_high_and_batch(struct zone *zone, | |
| 4266 | + struct per_cpu_pageset *pcp) | |
| 4266 | 4267 | { |
| 4267 | 4268 | if (percpu_pagelist_fraction) |
| 4268 | 4269 | pageset_set_high(pcp, |
| @@ -5881,23 +5882,38 @@ int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *table, int write, · 1612 unmodified lines | ||
| 5881 | 5882 | void __user *buffer, size_t *length, loff_t *ppos) |
| 5882 | 5883 | { |
| 5883 | 5884 | struct zone *zone; |
| 5884 | − unsigned int cpu; | |
| 5885 | + int old_percpu_pagelist_fraction; | |
| 5885 | 5886 | int ret; |
| 5886 | 5887 | |
| 5888 | + mutex_lock(&pcp_batch_high_lock); | |
| 5889 | + old_percpu_pagelist_fraction = percpu_pagelist_fraction; | |
| 5890 | + | |
| 5887 | 5891 | ret = proc_dointvec_minmax(table, write, buffer, length, ppos); |
| 5888 | − if (!write || (ret < 0)) | |
| 5889 | − return ret; | |
| 5892 | + if (!write || ret < 0) | |
| 5893 | + goto out; | |
| 5894 | + | |
| 5895 | + /* Sanity checking to avoid pcp imbalance */ | |
| 5896 | + if (percpu_pagelist_fraction && | |
| 5897 | + percpu_pagelist_fraction < MIN_PERCPU_PAGELIST_FRACTION) { | |
| 5898 | + percpu_pagelist_fraction = old_percpu_pagelist_fraction; | |
| 5899 | + ret = -EINVAL; | |
| 5900 | + goto out; | |
| 5901 | + } | |
| 5902 | + | |
| 5903 | + /* No change? */ | |
| 5904 | + if (percpu_pagelist_fraction == old_percpu_pagelist_fraction) | |
| 5905 | + goto out; | |
| 5890 | 5906 | |
| 5891 | − mutex_lock(&pcp_batch_high_lock); | |
| 5892 | 5907 | for_each_populated_zone(zone) { |
| 5893 | − unsigned long high; | |
| 5894 | − high = zone->managed_pages / percpu_pagelist_fraction; | |
| 5908 | + unsigned int cpu; | |
| 5909 | + | |
| 5895 | 5910 | for_each_possible_cpu(cpu) |
| 5896 | − pageset_set_high(per_cpu_ptr(zone->pageset, cpu), | |
| 5897 | − high); | |
| 5911 | + pageset_set_high_and_batch(zone, | |
| 5912 | + per_cpu_ptr(zone->pageset, cpu)); | |
| 5898 | 5913 | } |
| 5914 | +out: | |
| 5899 | 5915 | mutex_unlock(&pcp_batch_high_lock); |
| 5900 | − return 0; | |
| 5916 | + return ret; | |
| 5901 | 5917 | } |
| 5902 | 5918 | |
| 5903 | 5919 | int hashdist = HASHDIST_DEFAULT; |