There are several storage drivers like dm-multipath, iscsi, tcmu-runner,
amd nbd that have userspace components that can run in the IO path. For
example, iscsi and nbd's userspace deamons may need to recreate a socket
and/or send IO on it, and dm-multipath's daemon multipathd may need to
send SG IO or read/write IO to figure out the state of paths and re-set
them up.
In the kernel these drivers have access to GFP_NOIO/GFP_NOFS and the
memalloc_*_save/restore functions to control the allocation behavior,
but for userspace we would end up hitting an allocation that ended up
writing data back to the same device we are trying to allocate for.
The device is then in a state of deadlock, because to execute IO the
device needs to allocate memory, but to allocate memory the memory
layers want execute IO to the device.
Here is an example with nbd using a local userspace daemon that performs
network IO to a remote server. We are using XFS on top of the nbd device,
but it can happen with any FS or other modules layered on top of the nbd
device that can write out data to free memory. Here a nbd daemon helper
thread, msgr-worker-1, is performing a write/sendmsg on a socket to execute
a request. This kicks off a reclaim operation which results in a WRITE to
the nbd device and the nbd thread calling back into the mm layer.
[ 1626.609191] msgr-worker-1 D 0 1026 1 0x00004000
[ 1626.609193] Call Trace:
[ 1626.609195] ? __schedule+0x29b/0x630
[ 1626.609197] ? wait_for_completion+0xe0/0x170
[ 1626.609198] schedule+0x30/0xb0
[ 1626.609200] schedule_timeout+0x1f6/0x2f0
[ 1626.609202] ? blk_finish_plug+0x21/0x2e
[ 1626.609204] ? _xfs_buf_ioapply+0x2e6/0x410
[ 1626.609206] ? wait_for_completion+0xe0/0x170
[ 1626.609208] wait_for_completion+0x108/0x170
[ 1626.609210] ? wake_up_q+0x70/0x70
[ 1626.609212] ? __xfs_buf_submit+0x12e/0x250
[ 1626.609214] ? xfs_bwrite+0x25/0x60
[ 1626.609215] xfs_buf_iowait+0x22/0xf0
[ 1626.609218] __xfs_buf_submit+0x12e/0x250
[ 1626.609220] xfs_bwrite+0x25/0x60
[ 1626.609222] xfs_reclaim_inode+0x2e8/0x310
[ 1626.609224] xfs_reclaim_inodes_ag+0x1b6/0x300
[ 1626.609227] xfs_reclaim_inodes_nr+0x31/0x40
[ 1626.609228] super_cache_scan+0x152/0x1a0
[ 1626.609231] do_shrink_slab+0x12c/0x2d0
[ 1626.609233] shrink_slab+0x9c/0x2a0
[ 1626.609235] shrink_node+0xd7/0x470
[ 1626.609237] do_try_to_free_pages+0xbf/0x380
[ 1626.609240] try_to_free_pages+0xd9/0x1f0
[ 1626.609245] __alloc_pages_slowpath+0x3a4/0xd30
[ 1626.609251] ? ___slab_alloc+0x238/0x560
[ 1626.609254] __alloc_pages_nodemask+0x30c/0x350
[ 1626.609259] skb_page_frag_refill+0x97/0xd0
[ 1626.609274] sk_page_frag_refill+0x1d/0x80
[ 1626.609279] tcp_sendmsg_locked+0x2bb/0xdd0
[ 1626.609304] tcp_sendmsg+0x27/0x40
[ 1626.609307] sock_sendmsg+0x54/0x60
[ 1626.609308] ___sys_sendmsg+0x29f/0x320
[ 1626.609313] ? sock_poll+0x66/0xb0
[ 1626.609318] ? ep_item_poll.isra.15+0x40/0xc0
[ 1626.609320] ? ep_send_events_proc+0xe6/0x230
[ 1626.609322] ? hrtimer_try_to_cancel+0x54/0xf0
[ 1626.609324] ? ep_read_events_proc+0xc0/0xc0
[ 1626.609326] ? _raw_write_unlock_irq+0xa/0x20
[ 1626.609327] ? ep_scan_ready_list.constprop.19+0x218/0x230
[ 1626.609329] ? __hrtimer_init+0xb0/0xb0
[ 1626.609331] ? _raw_spin_unlock_irq+0xa/0x20
[ 1626.609334] ? ep_poll+0x26c/0x4a0
[ 1626.609337] ? tcp_tsq_write.part.54+0xa0/0xa0
[ 1626.609339] ? release_sock+0x43/0x90
[ 1626.609341] ? _raw_spin_unlock_bh+0xa/0x20
[ 1626.609342] __sys_sendmsg+0x47/0x80
[ 1626.609347] do_syscall_64+0x5f/0x1c0
[ 1626.609349] ? prepare_exit_to_usermode+0x75/0xa0
[ 1626.609351] entry_SYSCALL_64_after_hwframe+0x44/0xa9
This patch adds a new prctl command that daemons can use after they have
done their initial setup, and before they start to do allocations that
are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
flags so both userspace block and FS threads can use it to avoid the
allocation recursion and try to prevent from being throttled while
writing out data to free up memory.
Signed-off-by: Mike Christie <[email protected]>
Acked-by: Michal Hocko <[email protected]>
Tested-by: Masato Suzuki <[email protected]>
Reviewed-by: Damien Le Moal <[email protected]>
---
V4:
- Fix PR_GET_IO_FLUSHER check to match SET.
V3:
- Drop NOFS, set PF_LESS_THROTTLE and rename prctl flag to reflect it
is more general and can support both FS and block devices. Both fuse
and block device daemons, nbd and tcmu-runner, have been tested to
confirm the more restrictive PF_MEMALLOC_NOIO also works for fuse.
- Use CAP_SYS_RESOURCE instead of admin.
V2:
- Use prctl instead of procfs.
- Add support for NOFS for fuse.
- Check permissions.
include/uapi/linux/capability.h | 1 +
include/uapi/linux/prctl.h | 4 ++++
kernel/sys.c | 25 +++++++++++++++++++++++++
3 files changed, 30 insertions(+)
diff --git a/include/uapi/linux/capability.h b/include/uapi/linux/capability.h
index 240fdb9a60f6..272dc69fa080 100644
--- a/include/uapi/linux/capability.h
+++ b/include/uapi/linux/capability.h
@@ -301,6 +301,7 @@ struct vfs_ns_cap_data {
/* Allow more than 64hz interrupts from the real-time clock */
/* Override max number of consoles on console allocation */
/* Override max number of keymaps */
+/* Control memory reclaim behavior */
#define CAP_SYS_RESOURCE 24
diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
index 7da1b37b27aa..07b4f8131e36 100644
--- a/include/uapi/linux/prctl.h
+++ b/include/uapi/linux/prctl.h
@@ -234,4 +234,8 @@ struct prctl_mm_map {
#define PR_GET_TAGGED_ADDR_CTRL 56
# define PR_TAGGED_ADDR_ENABLE (1UL << 0)
+/* Control reclaim behavior when allocating memory */
+#define PR_SET_IO_FLUSHER 57
+#define PR_GET_IO_FLUSHER 58
+
#endif /* _LINUX_PRCTL_H */
diff --git a/kernel/sys.c b/kernel/sys.c
index a611d1d58c7d..c1a360370d09 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -2259,6 +2259,8 @@ int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
return -EINVAL;
}
+#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LESS_THROTTLE)
+
SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
unsigned long, arg4, unsigned long, arg5)
{
@@ -2486,6 +2488,29 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
return -EINVAL;
error = GET_TAGGED_ADDR_CTRL();
break;
+ case PR_SET_IO_FLUSHER:
+ if (!capable(CAP_SYS_RESOURCE))
+ return -EPERM;
+
+ if (arg3 || arg4 || arg5)
+ return -EINVAL;
+
+ if (arg2 == 1)
+ current->flags |= PR_IO_FLUSHER;
+ else if (!arg2)
+ current->flags &= ~PR_IO_FLUSHER;
+ else
+ return -EINVAL;
+ break;
+ case PR_GET_IO_FLUSHER:
+ if (!capable(CAP_SYS_RESOURCE))
+ return -EPERM;
+
+ if (arg2 || arg3 || arg4 || arg5)
+ return -EINVAL;
+
+ error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
+ break;
default:
error = -EINVAL;
break;
--
2.20.1
On 11/11/19 4:19 PM, Mike Christie wrote:
> There are several storage drivers like dm-multipath, iscsi, tcmu-runner,
> amd nbd that have userspace components that can run in the IO path. For
> example, iscsi and nbd's userspace deamons may need to recreate a socket
> and/or send IO on it, and dm-multipath's daemon multipathd may need to
> send SG IO or read/write IO to figure out the state of paths and re-set
> them up.
Reviewed-by: Bart Van Assche <[email protected]>
On Mon, Nov 11, 2019 at 4:19 PM Mike Christie <[email protected]> wrote:
>
> There are several storage drivers like dm-multipath, iscsi, tcmu-runner,
> amd nbd that have userspace components that can run in the IO path. For
> example, iscsi and nbd's userspace deamons may need to recreate a socket
> and/or send IO on it, and dm-multipath's daemon multipathd may need to
> send SG IO or read/write IO to figure out the state of paths and re-set
> them up.
>
> In the kernel these drivers have access to GFP_NOIO/GFP_NOFS and the
> memalloc_*_save/restore functions to control the allocation behavior,
> but for userspace we would end up hitting an allocation that ended up
> writing data back to the same device we are trying to allocate for.
> The device is then in a state of deadlock, because to execute IO the
> device needs to allocate memory, but to allocate memory the memory
> layers want execute IO to the device.
>
> Here is an example with nbd using a local userspace daemon that performs
> network IO to a remote server. We are using XFS on top of the nbd device,
> but it can happen with any FS or other modules layered on top of the nbd
> device that can write out data to free memory. Here a nbd daemon helper
> thread, msgr-worker-1, is performing a write/sendmsg on a socket to execute
> a request. This kicks off a reclaim operation which results in a WRITE to
> the nbd device and the nbd thread calling back into the mm layer.
>
> [ 1626.609191] msgr-worker-1 D 0 1026 1 0x00004000
> [ 1626.609193] Call Trace:
> [ 1626.609195] ? __schedule+0x29b/0x630
> [ 1626.609197] ? wait_for_completion+0xe0/0x170
> [ 1626.609198] schedule+0x30/0xb0
> [ 1626.609200] schedule_timeout+0x1f6/0x2f0
> [ 1626.609202] ? blk_finish_plug+0x21/0x2e
> [ 1626.609204] ? _xfs_buf_ioapply+0x2e6/0x410
> [ 1626.609206] ? wait_for_completion+0xe0/0x170
> [ 1626.609208] wait_for_completion+0x108/0x170
> [ 1626.609210] ? wake_up_q+0x70/0x70
> [ 1626.609212] ? __xfs_buf_submit+0x12e/0x250
> [ 1626.609214] ? xfs_bwrite+0x25/0x60
> [ 1626.609215] xfs_buf_iowait+0x22/0xf0
> [ 1626.609218] __xfs_buf_submit+0x12e/0x250
> [ 1626.609220] xfs_bwrite+0x25/0x60
> [ 1626.609222] xfs_reclaim_inode+0x2e8/0x310
> [ 1626.609224] xfs_reclaim_inodes_ag+0x1b6/0x300
> [ 1626.609227] xfs_reclaim_inodes_nr+0x31/0x40
> [ 1626.609228] super_cache_scan+0x152/0x1a0
> [ 1626.609231] do_shrink_slab+0x12c/0x2d0
> [ 1626.609233] shrink_slab+0x9c/0x2a0
> [ 1626.609235] shrink_node+0xd7/0x470
> [ 1626.609237] do_try_to_free_pages+0xbf/0x380
> [ 1626.609240] try_to_free_pages+0xd9/0x1f0
> [ 1626.609245] __alloc_pages_slowpath+0x3a4/0xd30
> [ 1626.609251] ? ___slab_alloc+0x238/0x560
> [ 1626.609254] __alloc_pages_nodemask+0x30c/0x350
> [ 1626.609259] skb_page_frag_refill+0x97/0xd0
> [ 1626.609274] sk_page_frag_refill+0x1d/0x80
> [ 1626.609279] tcp_sendmsg_locked+0x2bb/0xdd0
> [ 1626.609304] tcp_sendmsg+0x27/0x40
> [ 1626.609307] sock_sendmsg+0x54/0x60
> [ 1626.609308] ___sys_sendmsg+0x29f/0x320
> [ 1626.609313] ? sock_poll+0x66/0xb0
> [ 1626.609318] ? ep_item_poll.isra.15+0x40/0xc0
> [ 1626.609320] ? ep_send_events_proc+0xe6/0x230
> [ 1626.609322] ? hrtimer_try_to_cancel+0x54/0xf0
> [ 1626.609324] ? ep_read_events_proc+0xc0/0xc0
> [ 1626.609326] ? _raw_write_unlock_irq+0xa/0x20
> [ 1626.609327] ? ep_scan_ready_list.constprop.19+0x218/0x230
> [ 1626.609329] ? __hrtimer_init+0xb0/0xb0
> [ 1626.609331] ? _raw_spin_unlock_irq+0xa/0x20
> [ 1626.609334] ? ep_poll+0x26c/0x4a0
> [ 1626.609337] ? tcp_tsq_write.part.54+0xa0/0xa0
> [ 1626.609339] ? release_sock+0x43/0x90
> [ 1626.609341] ? _raw_spin_unlock_bh+0xa/0x20
> [ 1626.609342] __sys_sendmsg+0x47/0x80
> [ 1626.609347] do_syscall_64+0x5f/0x1c0
> [ 1626.609349] ? prepare_exit_to_usermode+0x75/0xa0
> [ 1626.609351] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>
> This patch adds a new prctl command that daemons can use after they have
> done their initial setup, and before they start to do allocations that
> are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
> flags so both userspace block and FS threads can use it to avoid the
> allocation recursion and try to prevent from being throttled while
> writing out data to free up memory.
>
> Signed-off-by: Mike Christie <[email protected]>
> Acked-by: Michal Hocko <[email protected]>
> Tested-by: Masato Suzuki <[email protected]>
> Reviewed-by: Damien Le Moal <[email protected]>
I suppose this patch should be routed through MM tree, so, CCing Andrew.
>
> ---
>
> V4:
> - Fix PR_GET_IO_FLUSHER check to match SET.
>
> V3:
> - Drop NOFS, set PF_LESS_THROTTLE and rename prctl flag to reflect it
> is more general and can support both FS and block devices. Both fuse
> and block device daemons, nbd and tcmu-runner, have been tested to
> confirm the more restrictive PF_MEMALLOC_NOIO also works for fuse.
>
> - Use CAP_SYS_RESOURCE instead of admin.
>
> V2:
> - Use prctl instead of procfs.
> - Add support for NOFS for fuse.
> - Check permissions.
>
>
> include/uapi/linux/capability.h | 1 +
> include/uapi/linux/prctl.h | 4 ++++
> kernel/sys.c | 25 +++++++++++++++++++++++++
> 3 files changed, 30 insertions(+)
>
> diff --git a/include/uapi/linux/capability.h b/include/uapi/linux/capability.h
> index 240fdb9a60f6..272dc69fa080 100644
> --- a/include/uapi/linux/capability.h
> +++ b/include/uapi/linux/capability.h
> @@ -301,6 +301,7 @@ struct vfs_ns_cap_data {
> /* Allow more than 64hz interrupts from the real-time clock */
> /* Override max number of consoles on console allocation */
> /* Override max number of keymaps */
> +/* Control memory reclaim behavior */
>
> #define CAP_SYS_RESOURCE 24
>
> diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
> index 7da1b37b27aa..07b4f8131e36 100644
> --- a/include/uapi/linux/prctl.h
> +++ b/include/uapi/linux/prctl.h
> @@ -234,4 +234,8 @@ struct prctl_mm_map {
> #define PR_GET_TAGGED_ADDR_CTRL 56
> # define PR_TAGGED_ADDR_ENABLE (1UL << 0)
>
> +/* Control reclaim behavior when allocating memory */
> +#define PR_SET_IO_FLUSHER 57
> +#define PR_GET_IO_FLUSHER 58
> +
> #endif /* _LINUX_PRCTL_H */
> diff --git a/kernel/sys.c b/kernel/sys.c
> index a611d1d58c7d..c1a360370d09 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -2259,6 +2259,8 @@ int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
> return -EINVAL;
> }
>
> +#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LESS_THROTTLE)
> +
> SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> unsigned long, arg4, unsigned long, arg5)
> {
> @@ -2486,6 +2488,29 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> return -EINVAL;
> error = GET_TAGGED_ADDR_CTRL();
> break;
> + case PR_SET_IO_FLUSHER:
> + if (!capable(CAP_SYS_RESOURCE))
> + return -EPERM;
> +
> + if (arg3 || arg4 || arg5)
> + return -EINVAL;
> +
> + if (arg2 == 1)
> + current->flags |= PR_IO_FLUSHER;
> + else if (!arg2)
> + current->flags &= ~PR_IO_FLUSHER;
> + else
> + return -EINVAL;
> + break;
> + case PR_GET_IO_FLUSHER:
> + if (!capable(CAP_SYS_RESOURCE))
> + return -EPERM;
> +
> + if (arg2 || arg3 || arg4 || arg5)
> + return -EINVAL;
> +
> + error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
> + break;
> default:
> error = -EINVAL;
> break;
> --
> 2.20.1
>
On 12/05/2019 04:43 PM, Shakeel Butt wrote:
> On Mon, Nov 11, 2019 at 4:19 PM Mike Christie <[email protected]> wrote:
>>
>> There are several storage drivers like dm-multipath, iscsi, tcmu-runner,
>> amd nbd that have userspace components that can run in the IO path. For
>> example, iscsi and nbd's userspace deamons may need to recreate a socket
>> and/or send IO on it, and dm-multipath's daemon multipathd may need to
>> send SG IO or read/write IO to figure out the state of paths and re-set
>> them up.
>>
>> In the kernel these drivers have access to GFP_NOIO/GFP_NOFS and the
>> memalloc_*_save/restore functions to control the allocation behavior,
>> but for userspace we would end up hitting an allocation that ended up
>> writing data back to the same device we are trying to allocate for.
>> The device is then in a state of deadlock, because to execute IO the
>> device needs to allocate memory, but to allocate memory the memory
>> layers want execute IO to the device.
>>
>> Here is an example with nbd using a local userspace daemon that performs
>> network IO to a remote server. We are using XFS on top of the nbd device,
>> but it can happen with any FS or other modules layered on top of the nbd
>> device that can write out data to free memory. Here a nbd daemon helper
>> thread, msgr-worker-1, is performing a write/sendmsg on a socket to execute
>> a request. This kicks off a reclaim operation which results in a WRITE to
>> the nbd device and the nbd thread calling back into the mm layer.
>>
>> [ 1626.609191] msgr-worker-1 D 0 1026 1 0x00004000
>> [ 1626.609193] Call Trace:
>> [ 1626.609195] ? __schedule+0x29b/0x630
>> [ 1626.609197] ? wait_for_completion+0xe0/0x170
>> [ 1626.609198] schedule+0x30/0xb0
>> [ 1626.609200] schedule_timeout+0x1f6/0x2f0
>> [ 1626.609202] ? blk_finish_plug+0x21/0x2e
>> [ 1626.609204] ? _xfs_buf_ioapply+0x2e6/0x410
>> [ 1626.609206] ? wait_for_completion+0xe0/0x170
>> [ 1626.609208] wait_for_completion+0x108/0x170
>> [ 1626.609210] ? wake_up_q+0x70/0x70
>> [ 1626.609212] ? __xfs_buf_submit+0x12e/0x250
>> [ 1626.609214] ? xfs_bwrite+0x25/0x60
>> [ 1626.609215] xfs_buf_iowait+0x22/0xf0
>> [ 1626.609218] __xfs_buf_submit+0x12e/0x250
>> [ 1626.609220] xfs_bwrite+0x25/0x60
>> [ 1626.609222] xfs_reclaim_inode+0x2e8/0x310
>> [ 1626.609224] xfs_reclaim_inodes_ag+0x1b6/0x300
>> [ 1626.609227] xfs_reclaim_inodes_nr+0x31/0x40
>> [ 1626.609228] super_cache_scan+0x152/0x1a0
>> [ 1626.609231] do_shrink_slab+0x12c/0x2d0
>> [ 1626.609233] shrink_slab+0x9c/0x2a0
>> [ 1626.609235] shrink_node+0xd7/0x470
>> [ 1626.609237] do_try_to_free_pages+0xbf/0x380
>> [ 1626.609240] try_to_free_pages+0xd9/0x1f0
>> [ 1626.609245] __alloc_pages_slowpath+0x3a4/0xd30
>> [ 1626.609251] ? ___slab_alloc+0x238/0x560
>> [ 1626.609254] __alloc_pages_nodemask+0x30c/0x350
>> [ 1626.609259] skb_page_frag_refill+0x97/0xd0
>> [ 1626.609274] sk_page_frag_refill+0x1d/0x80
>> [ 1626.609279] tcp_sendmsg_locked+0x2bb/0xdd0
>> [ 1626.609304] tcp_sendmsg+0x27/0x40
>> [ 1626.609307] sock_sendmsg+0x54/0x60
>> [ 1626.609308] ___sys_sendmsg+0x29f/0x320
>> [ 1626.609313] ? sock_poll+0x66/0xb0
>> [ 1626.609318] ? ep_item_poll.isra.15+0x40/0xc0
>> [ 1626.609320] ? ep_send_events_proc+0xe6/0x230
>> [ 1626.609322] ? hrtimer_try_to_cancel+0x54/0xf0
>> [ 1626.609324] ? ep_read_events_proc+0xc0/0xc0
>> [ 1626.609326] ? _raw_write_unlock_irq+0xa/0x20
>> [ 1626.609327] ? ep_scan_ready_list.constprop.19+0x218/0x230
>> [ 1626.609329] ? __hrtimer_init+0xb0/0xb0
>> [ 1626.609331] ? _raw_spin_unlock_irq+0xa/0x20
>> [ 1626.609334] ? ep_poll+0x26c/0x4a0
>> [ 1626.609337] ? tcp_tsq_write.part.54+0xa0/0xa0
>> [ 1626.609339] ? release_sock+0x43/0x90
>> [ 1626.609341] ? _raw_spin_unlock_bh+0xa/0x20
>> [ 1626.609342] __sys_sendmsg+0x47/0x80
>> [ 1626.609347] do_syscall_64+0x5f/0x1c0
>> [ 1626.609349] ? prepare_exit_to_usermode+0x75/0xa0
>> [ 1626.609351] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>>
>> This patch adds a new prctl command that daemons can use after they have
>> done their initial setup, and before they start to do allocations that
>> are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
>> flags so both userspace block and FS threads can use it to avoid the
>> allocation recursion and try to prevent from being throttled while
>> writing out data to free up memory.
>>
>> Signed-off-by: Mike Christie <[email protected]>
>> Acked-by: Michal Hocko <[email protected]>
>> Tested-by: Masato Suzuki <[email protected]>
>> Reviewed-by: Damien Le Moal <[email protected]>
>
> I suppose this patch should be routed through MM tree, so, CCing Andrew.
>
Andrew and other mm/storage developers,
Do I need to handle anything else for this patch, or are there any other
concerns? Is this maybe something we want to talk about at a quick LSF
session?
I have retested it with Linus's current tree. It still applies cleanly
(just some offsets), and fixes the problem described above we have been
hitting.
On Fri, Jan 24, 2020 at 10:22:33AM -0600, Mike Christie wrote:
> On 12/05/2019 04:43 PM, Shakeel Butt wrote:
> > On Mon, Nov 11, 2019 at 4:19 PM Mike Christie <[email protected]> wrote:
> >> This patch adds a new prctl command that daemons can use after they have
> >> done their initial setup, and before they start to do allocations that
> >> are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
> >> flags so both userspace block and FS threads can use it to avoid the
> >> allocation recursion and try to prevent from being throttled while
> >> writing out data to free up memory.
> >>
> >> Signed-off-by: Mike Christie <[email protected]>
> >> Acked-by: Michal Hocko <[email protected]>
> >> Tested-by: Masato Suzuki <[email protected]>
> >> Reviewed-by: Damien Le Moal <[email protected]>
> >
> > I suppose this patch should be routed through MM tree, so, CCing Andrew.
> >
>
> Andrew and other mm/storage developers,
>
> Do I need to handle anything else for this patch, or are there any other
> concerns? Is this maybe something we want to talk about at a quick LSF
> session?
>
> I have retested it with Linus's current tree. It still applies cleanly
> (just some offsets), and fixes the problem described above we have been
> hitting.
I must have missed this version being posted (just looked it up on
lore.kernel.org). As far as I'm concerned this is good to go and it
is absolutely necessary for userspace IO stacks to function
correctly.
Reviewed-by: Dave Chinner <[email protected]>
If no manintainer picks it up before the next merge window, then I
recommend resending the latest version to Linus asking him to merge
it.
Cheers,
Dave.
--
Dave Chinner
[email protected]
On Mon, Nov 11, 2019 at 06:19:00PM -0600, Mike Christie wrote:
> There are several storage drivers like dm-multipath, iscsi, tcmu-runner,
> amd nbd that have userspace components that can run in the IO path. For
> example, iscsi and nbd's userspace deamons may need to recreate a socket
> and/or send IO on it, and dm-multipath's daemon multipathd may need to
> send SG IO or read/write IO to figure out the state of paths and re-set
> them up.
>
> In the kernel these drivers have access to GFP_NOIO/GFP_NOFS and the
> memalloc_*_save/restore functions to control the allocation behavior,
> but for userspace we would end up hitting an allocation that ended up
> writing data back to the same device we are trying to allocate for.
> The device is then in a state of deadlock, because to execute IO the
> device needs to allocate memory, but to allocate memory the memory
> layers want execute IO to the device.
>
> Here is an example with nbd using a local userspace daemon that performs
> network IO to a remote server. We are using XFS on top of the nbd device,
> but it can happen with any FS or other modules layered on top of the nbd
> device that can write out data to free memory. Here a nbd daemon helper
> thread, msgr-worker-1, is performing a write/sendmsg on a socket to execute
> a request. This kicks off a reclaim operation which results in a WRITE to
> the nbd device and the nbd thread calling back into the mm layer.
>
/me would like to see the documentation update to prctl(2).
Assuming that "set takes 0 or 1, get returns 0 or 1" is more or less how
this interface is supposed to work,
Reviewed-by: Darrick J. Wong <[email protected]>
--D
> [ 1626.609191] msgr-worker-1 D 0 1026 1 0x00004000
> [ 1626.609193] Call Trace:
> [ 1626.609195] ? __schedule+0x29b/0x630
> [ 1626.609197] ? wait_for_completion+0xe0/0x170
> [ 1626.609198] schedule+0x30/0xb0
> [ 1626.609200] schedule_timeout+0x1f6/0x2f0
> [ 1626.609202] ? blk_finish_plug+0x21/0x2e
> [ 1626.609204] ? _xfs_buf_ioapply+0x2e6/0x410
> [ 1626.609206] ? wait_for_completion+0xe0/0x170
> [ 1626.609208] wait_for_completion+0x108/0x170
> [ 1626.609210] ? wake_up_q+0x70/0x70
> [ 1626.609212] ? __xfs_buf_submit+0x12e/0x250
> [ 1626.609214] ? xfs_bwrite+0x25/0x60
> [ 1626.609215] xfs_buf_iowait+0x22/0xf0
> [ 1626.609218] __xfs_buf_submit+0x12e/0x250
> [ 1626.609220] xfs_bwrite+0x25/0x60
> [ 1626.609222] xfs_reclaim_inode+0x2e8/0x310
> [ 1626.609224] xfs_reclaim_inodes_ag+0x1b6/0x300
> [ 1626.609227] xfs_reclaim_inodes_nr+0x31/0x40
> [ 1626.609228] super_cache_scan+0x152/0x1a0
> [ 1626.609231] do_shrink_slab+0x12c/0x2d0
> [ 1626.609233] shrink_slab+0x9c/0x2a0
> [ 1626.609235] shrink_node+0xd7/0x470
> [ 1626.609237] do_try_to_free_pages+0xbf/0x380
> [ 1626.609240] try_to_free_pages+0xd9/0x1f0
> [ 1626.609245] __alloc_pages_slowpath+0x3a4/0xd30
> [ 1626.609251] ? ___slab_alloc+0x238/0x560
> [ 1626.609254] __alloc_pages_nodemask+0x30c/0x350
> [ 1626.609259] skb_page_frag_refill+0x97/0xd0
> [ 1626.609274] sk_page_frag_refill+0x1d/0x80
> [ 1626.609279] tcp_sendmsg_locked+0x2bb/0xdd0
> [ 1626.609304] tcp_sendmsg+0x27/0x40
> [ 1626.609307] sock_sendmsg+0x54/0x60
> [ 1626.609308] ___sys_sendmsg+0x29f/0x320
> [ 1626.609313] ? sock_poll+0x66/0xb0
> [ 1626.609318] ? ep_item_poll.isra.15+0x40/0xc0
> [ 1626.609320] ? ep_send_events_proc+0xe6/0x230
> [ 1626.609322] ? hrtimer_try_to_cancel+0x54/0xf0
> [ 1626.609324] ? ep_read_events_proc+0xc0/0xc0
> [ 1626.609326] ? _raw_write_unlock_irq+0xa/0x20
> [ 1626.609327] ? ep_scan_ready_list.constprop.19+0x218/0x230
> [ 1626.609329] ? __hrtimer_init+0xb0/0xb0
> [ 1626.609331] ? _raw_spin_unlock_irq+0xa/0x20
> [ 1626.609334] ? ep_poll+0x26c/0x4a0
> [ 1626.609337] ? tcp_tsq_write.part.54+0xa0/0xa0
> [ 1626.609339] ? release_sock+0x43/0x90
> [ 1626.609341] ? _raw_spin_unlock_bh+0xa/0x20
> [ 1626.609342] __sys_sendmsg+0x47/0x80
> [ 1626.609347] do_syscall_64+0x5f/0x1c0
> [ 1626.609349] ? prepare_exit_to_usermode+0x75/0xa0
> [ 1626.609351] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>
> This patch adds a new prctl command that daemons can use after they have
> done their initial setup, and before they start to do allocations that
> are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
> flags so both userspace block and FS threads can use it to avoid the
> allocation recursion and try to prevent from being throttled while
> writing out data to free up memory.
>
> Signed-off-by: Mike Christie <[email protected]>
> Acked-by: Michal Hocko <[email protected]>
> Tested-by: Masato Suzuki <[email protected]>
> Reviewed-by: Damien Le Moal <[email protected]>
>
> ---
>
> V4:
> - Fix PR_GET_IO_FLUSHER check to match SET.
>
> V3:
> - Drop NOFS, set PF_LESS_THROTTLE and rename prctl flag to reflect it
> is more general and can support both FS and block devices. Both fuse
> and block device daemons, nbd and tcmu-runner, have been tested to
> confirm the more restrictive PF_MEMALLOC_NOIO also works for fuse.
>
> - Use CAP_SYS_RESOURCE instead of admin.
>
> V2:
> - Use prctl instead of procfs.
> - Add support for NOFS for fuse.
> - Check permissions.
>
>
> include/uapi/linux/capability.h | 1 +
> include/uapi/linux/prctl.h | 4 ++++
> kernel/sys.c | 25 +++++++++++++++++++++++++
> 3 files changed, 30 insertions(+)
>
> diff --git a/include/uapi/linux/capability.h b/include/uapi/linux/capability.h
> index 240fdb9a60f6..272dc69fa080 100644
> --- a/include/uapi/linux/capability.h
> +++ b/include/uapi/linux/capability.h
> @@ -301,6 +301,7 @@ struct vfs_ns_cap_data {
> /* Allow more than 64hz interrupts from the real-time clock */
> /* Override max number of consoles on console allocation */
> /* Override max number of keymaps */
> +/* Control memory reclaim behavior */
>
> #define CAP_SYS_RESOURCE 24
>
> diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
> index 7da1b37b27aa..07b4f8131e36 100644
> --- a/include/uapi/linux/prctl.h
> +++ b/include/uapi/linux/prctl.h
> @@ -234,4 +234,8 @@ struct prctl_mm_map {
> #define PR_GET_TAGGED_ADDR_CTRL 56
> # define PR_TAGGED_ADDR_ENABLE (1UL << 0)
>
> +/* Control reclaim behavior when allocating memory */
> +#define PR_SET_IO_FLUSHER 57
> +#define PR_GET_IO_FLUSHER 58
> +
> #endif /* _LINUX_PRCTL_H */
> diff --git a/kernel/sys.c b/kernel/sys.c
> index a611d1d58c7d..c1a360370d09 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -2259,6 +2259,8 @@ int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
> return -EINVAL;
> }
>
> +#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LESS_THROTTLE)
> +
> SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> unsigned long, arg4, unsigned long, arg5)
> {
> @@ -2486,6 +2488,29 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> return -EINVAL;
> error = GET_TAGGED_ADDR_CTRL();
> break;
> + case PR_SET_IO_FLUSHER:
> + if (!capable(CAP_SYS_RESOURCE))
> + return -EPERM;
> +
> + if (arg3 || arg4 || arg5)
> + return -EINVAL;
> +
> + if (arg2 == 1)
> + current->flags |= PR_IO_FLUSHER;
> + else if (!arg2)
> + current->flags &= ~PR_IO_FLUSHER;
> + else
> + return -EINVAL;
> + break;
> + case PR_GET_IO_FLUSHER:
> + if (!capable(CAP_SYS_RESOURCE))
> + return -EPERM;
> +
> + if (arg2 || arg3 || arg4 || arg5)
> + return -EINVAL;
> +
> + error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
> + break;
> default:
> error = -EINVAL;
> break;
> --
> 2.20.1
>
On Sat, Jan 25, 2020 at 08:16:42AM +1100, Dave Chinner wrote:
> On Fri, Jan 24, 2020 at 10:22:33AM -0600, Mike Christie wrote:
> > On 12/05/2019 04:43 PM, Shakeel Butt wrote:
> > > On Mon, Nov 11, 2019 at 4:19 PM Mike Christie <[email protected]> wrote:
> > >> This patch adds a new prctl command that daemons can use after they have
> > >> done their initial setup, and before they start to do allocations that
> > >> are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
> > >> flags so both userspace block and FS threads can use it to avoid the
> > >> allocation recursion and try to prevent from being throttled while
> > >> writing out data to free up memory.
> > >>
> > >> Signed-off-by: Mike Christie <[email protected]>
> > >> Acked-by: Michal Hocko <[email protected]>
> > >> Tested-by: Masato Suzuki <[email protected]>
> > >> Reviewed-by: Damien Le Moal <[email protected]>
> > >
> > > I suppose this patch should be routed through MM tree, so, CCing Andrew.
> > >
> >
> > Andrew and other mm/storage developers,
> >
> > Do I need to handle anything else for this patch, or are there any other
> > concerns? Is this maybe something we want to talk about at a quick LSF
> > session?
> >
> > I have retested it with Linus's current tree. It still applies cleanly
> > (just some offsets), and fixes the problem described above we have been
> > hitting.
>
> I must have missed this version being posted (just looked it up on
> lore.kernel.org). As far as I'm concerned this is good to go and it
> is absolutely necessary for userspace IO stacks to function
> correctly.
>
> Reviewed-by: Dave Chinner <[email protected]>
>
> If no manintainer picks it up before the next merge window, then I
Since prctl() is thread-management and fs people seem to be happy and
have acked it I can pick this up too if noone objects and send this
along with the rest of process management.
Christian
On Mon, Jan 27, 2020 at 02:02:59PM +0100, Christian Brauner wrote:
> On Sat, Jan 25, 2020 at 08:16:42AM +1100, Dave Chinner wrote:
> > On Fri, Jan 24, 2020 at 10:22:33AM -0600, Mike Christie wrote:
> > > On 12/05/2019 04:43 PM, Shakeel Butt wrote:
> > > > On Mon, Nov 11, 2019 at 4:19 PM Mike Christie <[email protected]> wrote:
> > > >> This patch adds a new prctl command that daemons can use after they have
> > > >> done their initial setup, and before they start to do allocations that
> > > >> are in the IO path. It sets the PF_MEMALLOC_NOIO and PF_LESS_THROTTLE
> > > >> flags so both userspace block and FS threads can use it to avoid the
> > > >> allocation recursion and try to prevent from being throttled while
> > > >> writing out data to free up memory.
> > > >>
> > > >> Signed-off-by: Mike Christie <[email protected]>
> > > >> Acked-by: Michal Hocko <[email protected]>
> > > >> Tested-by: Masato Suzuki <[email protected]>
> > > >> Reviewed-by: Damien Le Moal <[email protected]>
> > > >
> > > > I suppose this patch should be routed through MM tree, so, CCing Andrew.
> > > >
> > >
> > > Andrew and other mm/storage developers,
> > >
> > > Do I need to handle anything else for this patch, or are there any other
> > > concerns? Is this maybe something we want to talk about at a quick LSF
> > > session?
> > >
> > > I have retested it with Linus's current tree. It still applies cleanly
> > > (just some offsets), and fixes the problem described above we have been
> > > hitting.
> >
> > I must have missed this version being posted (just looked it up on
> > lore.kernel.org). As far as I'm concerned this is good to go and it
> > is absolutely necessary for userspace IO stacks to function
> > correctly.
> >
> > Reviewed-by: Dave Chinner <[email protected]>
> >
> > If no manintainer picks it up before the next merge window, then I
>
> Since prctl() is thread-management and fs people seem to be happy and
> have acked it I can pick this up too if noone objects and send this
> along with the rest of process management.
This is upstream now
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8d19f1c8e1937baf74e1962aae9f90fa3aeab463
Christian