linux/drivers/misc/vmw_vmci/vmci_doorbell.c

605 lines
17 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* VMware VMCI Driver
*
* Copyright (C) 2012 VMware, Inc. All rights reserved.
*/
#include <linux/vmw_vmci_defs.h>
#include <linux/vmw_vmci_api.h>
#include <linux/completion.h>
#include <linux/hash.h>
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include "vmci_datagram.h"
#include "vmci_doorbell.h"
#include "vmci_resource.h"
#include "vmci_driver.h"
#include "vmci_route.h"
#define VMCI_DOORBELL_INDEX_BITS 6
#define VMCI_DOORBELL_INDEX_TABLE_SIZE (1 << VMCI_DOORBELL_INDEX_BITS)
#define VMCI_DOORBELL_HASH(_idx) hash_32(_idx, VMCI_DOORBELL_INDEX_BITS)
/*
* DoorbellEntry describes the a doorbell notification handle allocated by the
* host.
*/
struct dbell_entry {
struct vmci_resource resource;
struct hlist_node node;
struct work_struct work;
vmci_callback notify_cb;
void *client_data;
u32 idx;
u32 priv_flags;
bool run_delayed;
atomic_t active; /* Only used by guest personality */
};
/* The VMCI index table keeps track of currently registered doorbells. */
struct dbell_index_table {
spinlock_t lock; /* Index table lock */
struct hlist_head entries[VMCI_DOORBELL_INDEX_TABLE_SIZE];
};
static struct dbell_index_table vmci_doorbell_it = {
.lock = __SPIN_LOCK_UNLOCKED(vmci_doorbell_it.lock),
};
/*
* The max_notify_idx is one larger than the currently known bitmap index in
* use, and is used to determine how much of the bitmap needs to be scanned.
*/
static u32 max_notify_idx;
/*
* The notify_idx_count is used for determining whether there are free entries
* within the bitmap (if notify_idx_count + 1 < max_notify_idx).
*/
static u32 notify_idx_count;
/*
* The last_notify_idx_reserved is used to track the last index handed out - in
* the case where multiple handles share a notification index, we hand out
* indexes round robin based on last_notify_idx_reserved.
*/
static u32 last_notify_idx_reserved;
/* This is a one entry cache used to by the index allocation. */
static u32 last_notify_idx_released = PAGE_SIZE;
/*
* Utility function that retrieves the privilege flags associated
* with a given doorbell handle. For guest endpoints, the
* privileges are determined by the context ID, but for host
* endpoints privileges are associated with the complete
* handle. Hypervisor endpoints are not yet supported.
*/
int vmci_dbell_get_priv_flags(struct vmci_handle handle, u32 *priv_flags)
{
if (priv_flags == NULL || handle.context == VMCI_INVALID_ID)
return VMCI_ERROR_INVALID_ARGS;
if (handle.context == VMCI_HOST_CONTEXT_ID) {
struct dbell_entry *entry;
struct vmci_resource *resource;
resource = vmci_resource_by_handle(handle,
VMCI_RESOURCE_TYPE_DOORBELL);
if (!resource)
return VMCI_ERROR_NOT_FOUND;
entry = container_of(resource, struct dbell_entry, resource);
*priv_flags = entry->priv_flags;
vmci_resource_put(resource);
} else if (handle.context == VMCI_HYPERVISOR_CONTEXT_ID) {
/*
* Hypervisor endpoints for notifications are not
* supported (yet).
*/
return VMCI_ERROR_INVALID_ARGS;
} else {
*priv_flags = vmci_context_get_priv_flags(handle.context);
}
return VMCI_SUCCESS;
}
/*
* Find doorbell entry by bitmap index.
*/
static struct dbell_entry *dbell_index_table_find(u32 idx)
{
u32 bucket = VMCI_DOORBELL_HASH(idx);
struct dbell_entry *dbell;
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
hlist_for_each_entry(dbell, &vmci_doorbell_it.entries[bucket],
node) {
if (idx == dbell->idx)
return dbell;
}
return NULL;
}
/*
* Add the given entry to the index table. This willi take a reference to the
* entry's resource so that the entry is not deleted before it is removed from
* the * table.
*/
static void dbell_index_table_add(struct dbell_entry *entry)
{
u32 bucket;
u32 new_notify_idx;
vmci_resource_get(&entry->resource);
spin_lock_bh(&vmci_doorbell_it.lock);
/*
* Below we try to allocate an index in the notification
* bitmap with "not too much" sharing between resources. If we
* use less that the full bitmap, we either add to the end if
* there are no unused flags within the currently used area,
* or we search for unused ones. If we use the full bitmap, we
* allocate the index round robin.
*/
if (max_notify_idx < PAGE_SIZE || notify_idx_count < PAGE_SIZE) {
if (last_notify_idx_released < max_notify_idx &&
!dbell_index_table_find(last_notify_idx_released)) {
new_notify_idx = last_notify_idx_released;
last_notify_idx_released = PAGE_SIZE;
} else {
bool reused = false;
new_notify_idx = last_notify_idx_reserved;
if (notify_idx_count + 1 < max_notify_idx) {
do {
if (!dbell_index_table_find
(new_notify_idx)) {
reused = true;
break;
}
new_notify_idx = (new_notify_idx + 1) %
max_notify_idx;
} while (new_notify_idx !=
last_notify_idx_released);
}
if (!reused) {
new_notify_idx = max_notify_idx;
max_notify_idx++;
}
}
} else {
new_notify_idx = (last_notify_idx_reserved + 1) % PAGE_SIZE;
}
last_notify_idx_reserved = new_notify_idx;
notify_idx_count++;
entry->idx = new_notify_idx;
bucket = VMCI_DOORBELL_HASH(entry->idx);
hlist_add_head(&entry->node, &vmci_doorbell_it.entries[bucket]);
spin_unlock_bh(&vmci_doorbell_it.lock);
}
/*
* Remove the given entry from the index table. This will release() the
* entry's resource.
*/
static void dbell_index_table_remove(struct dbell_entry *entry)
{
spin_lock_bh(&vmci_doorbell_it.lock);
hlist_del_init(&entry->node);
notify_idx_count--;
if (entry->idx == max_notify_idx - 1) {
/*
* If we delete an entry with the maximum known
* notification index, we take the opportunity to
* prune the current max. As there might be other
* unused indices immediately below, we lower the
* maximum until we hit an index in use.
*/
while (max_notify_idx > 0 &&
!dbell_index_table_find(max_notify_idx - 1))
max_notify_idx--;
}
last_notify_idx_released = entry->idx;
spin_unlock_bh(&vmci_doorbell_it.lock);
vmci_resource_put(&entry->resource);
}
/*
* Creates a link between the given doorbell handle and the given
* index in the bitmap in the device backend. A notification state
* is created in hypervisor.
*/
static int dbell_link(struct vmci_handle handle, u32 notify_idx)
{
struct vmci_doorbell_link_msg link_msg;
link_msg.hdr.dst = vmci_make_handle(VMCI_HYPERVISOR_CONTEXT_ID,
VMCI_DOORBELL_LINK);
link_msg.hdr.src = VMCI_ANON_SRC_HANDLE;
link_msg.hdr.payload_size = sizeof(link_msg) - VMCI_DG_HEADERSIZE;
link_msg.handle = handle;
link_msg.notify_idx = notify_idx;
return vmci_send_datagram(&link_msg.hdr);
}
/*
* Unlinks the given doorbell handle from an index in the bitmap in
* the device backend. The notification state is destroyed in hypervisor.
*/
static int dbell_unlink(struct vmci_handle handle)
{
struct vmci_doorbell_unlink_msg unlink_msg;
unlink_msg.hdr.dst = vmci_make_handle(VMCI_HYPERVISOR_CONTEXT_ID,
VMCI_DOORBELL_UNLINK);
unlink_msg.hdr.src = VMCI_ANON_SRC_HANDLE;
unlink_msg.hdr.payload_size = sizeof(unlink_msg) - VMCI_DG_HEADERSIZE;
unlink_msg.handle = handle;
return vmci_send_datagram(&unlink_msg.hdr);
}
/*
* Notify another guest or the host. We send a datagram down to the
* host via the hypervisor with the notification info.
*/
static int dbell_notify_as_guest(struct vmci_handle handle, u32 priv_flags)
{
struct vmci_doorbell_notify_msg notify_msg;
notify_msg.hdr.dst = vmci_make_handle(VMCI_HYPERVISOR_CONTEXT_ID,
VMCI_DOORBELL_NOTIFY);
notify_msg.hdr.src = VMCI_ANON_SRC_HANDLE;
notify_msg.hdr.payload_size = sizeof(notify_msg) - VMCI_DG_HEADERSIZE;
notify_msg.handle = handle;
return vmci_send_datagram(&notify_msg.hdr);
}
/*
* Calls the specified callback in a delayed context.
*/
static void dbell_delayed_dispatch(struct work_struct *work)
{
struct dbell_entry *entry = container_of(work,
struct dbell_entry, work);
entry->notify_cb(entry->client_data);
vmci_resource_put(&entry->resource);
}
/*
* Dispatches a doorbell notification to the host context.
*/
int vmci_dbell_host_context_notify(u32 src_cid, struct vmci_handle handle)
{
struct dbell_entry *entry;
struct vmci_resource *resource;
if (vmci_handle_is_invalid(handle)) {
pr_devel("Notifying an invalid doorbell (handle=0x%x:0x%x)\n",
handle.context, handle.resource);
return VMCI_ERROR_INVALID_ARGS;
}
resource = vmci_resource_by_handle(handle,
VMCI_RESOURCE_TYPE_DOORBELL);
if (!resource) {
pr_devel("Notifying an unknown doorbell (handle=0x%x:0x%x)\n",
handle.context, handle.resource);
return VMCI_ERROR_NOT_FOUND;
}
entry = container_of(resource, struct dbell_entry, resource);
if (entry->run_delayed) {
VMCI: Release resource if the work is already queued Francois reported that VMware balloon gets stuck after a balloon reset, when the VMCI doorbell is removed. A similar error can occur when the balloon driver is removed with the following splat: [ 1088.622000] INFO: task modprobe:3565 blocked for more than 120 seconds. [ 1088.622035] Tainted: G W 5.2.0 #4 [ 1088.622087] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1088.622205] modprobe D 0 3565 1450 0x00000000 [ 1088.622210] Call Trace: [ 1088.622246] __schedule+0x2a8/0x690 [ 1088.622248] schedule+0x2d/0x90 [ 1088.622250] schedule_timeout+0x1d3/0x2f0 [ 1088.622252] wait_for_completion+0xba/0x140 [ 1088.622320] ? wake_up_q+0x80/0x80 [ 1088.622370] vmci_resource_remove+0xb9/0xc0 [vmw_vmci] [ 1088.622373] vmci_doorbell_destroy+0x9e/0xd0 [vmw_vmci] [ 1088.622379] vmballoon_vmci_cleanup+0x6e/0xf0 [vmw_balloon] [ 1088.622381] vmballoon_exit+0x18/0xcc8 [vmw_balloon] [ 1088.622394] __x64_sys_delete_module+0x146/0x280 [ 1088.622408] do_syscall_64+0x5a/0x130 [ 1088.622410] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 1088.622415] RIP: 0033:0x7f54f62791b7 [ 1088.622421] Code: Bad RIP value. [ 1088.622421] RSP: 002b:00007fff2a949008 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0 [ 1088.622426] RAX: ffffffffffffffda RBX: 000055dff8b55d00 RCX: 00007f54f62791b7 [ 1088.622426] RDX: 0000000000000000 RSI: 0000000000000800 RDI: 000055dff8b55d68 [ 1088.622427] RBP: 000055dff8b55d00 R08: 00007fff2a947fb1 R09: 0000000000000000 [ 1088.622427] R10: 00007f54f62f5cc0 R11: 0000000000000206 R12: 000055dff8b55d68 [ 1088.622428] R13: 0000000000000001 R14: 000055dff8b55d68 R15: 00007fff2a94a3f0 The cause for the bug is that when the "delayed" doorbell is invoked, it takes a reference on the doorbell entry and schedules work that is supposed to run the appropriate code and drop the doorbell entry reference. The code ignores the fact that if the work is already queued, it will not be scheduled to run one more time. As a result one of the references would not be dropped. When the code waits for the reference to get to zero, during balloon reset or module removal, it gets stuck. Fix it. Drop the reference if schedule_work() indicates that the work is already queued. Note that this bug got more apparent (or apparent at all) due to commit ce664331b248 ("vmw_balloon: VMCI_DOORBELL_SET does not check status"). Fixes: 83e2ec765be03 ("VMCI: doorbell implementation.") Reported-by: Francois Rigault <rigault.francois@gmail.com> Cc: Jorgen Hansen <jhansen@vmware.com> Cc: Adit Ranadive <aditr@vmware.com> Cc: Alexios Zavras <alexios.zavras@intel.com> Cc: Vishnu DASA <vdasa@vmware.com> Cc: stable@vger.kernel.org Signed-off-by: Nadav Amit <namit@vmware.com> Reviewed-by: Vishnu Dasa <vdasa@vmware.com> Link: https://lore.kernel.org/r/20190820202638.49003-1-namit@vmware.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-08-20 20:26:38 +00:00
if (!schedule_work(&entry->work))
vmci_resource_put(resource);
} else {
entry->notify_cb(entry->client_data);
vmci_resource_put(resource);
}
return VMCI_SUCCESS;
}
/*
* Register the notification bitmap with the host.
*/
bool vmci_dbell_register_notification_bitmap(u64 bitmap_ppn)
{
int result;
misc: vmw_vmci: explicitly initialize vmci_notify_bm_set_msg struct KMSAN complains that the vmci_use_ppn64() == false path in vmci_dbell_register_notification_bitmap() left upper 32bits of bitmap_set_msg.bitmap_ppn64 member uninitialized. ===================================================== BUG: KMSAN: uninit-value in kmsan_check_memory+0xd/0x10 CPU: 1 PID: 1 Comm: swapper/0 Not tainted 5.11.0-rc7+ #4 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 02/27/2020 Call Trace: dump_stack+0x21c/0x280 kmsan_report+0xfb/0x1e0 kmsan_internal_check_memory+0x484/0x520 kmsan_check_memory+0xd/0x10 iowrite8_rep+0x86/0x380 vmci_send_datagram+0x150/0x280 vmci_dbell_register_notification_bitmap+0x133/0x1e0 vmci_guest_probe_device+0xcab/0x1e70 pci_device_probe+0xab3/0xe70 really_probe+0xd16/0x24d0 driver_probe_device+0x29d/0x3a0 device_driver_attach+0x25a/0x490 __driver_attach+0x78c/0x840 bus_for_each_dev+0x210/0x340 driver_attach+0x89/0xb0 bus_add_driver+0x677/0xc40 driver_register+0x485/0x8e0 __pci_register_driver+0x1ff/0x350 vmci_guest_init+0x3e/0x41 vmci_drv_init+0x1d6/0x43f do_one_initcall+0x39c/0x9a0 do_initcall_level+0x1d7/0x259 do_initcalls+0x127/0x1cb do_basic_setup+0x33/0x36 kernel_init_freeable+0x29a/0x3ed kernel_init+0x1f/0x840 ret_from_fork+0x1f/0x30 Local variable ----bitmap_set_msg@vmci_dbell_register_notification_bitmap created at: vmci_dbell_register_notification_bitmap+0x50/0x1e0 vmci_dbell_register_notification_bitmap+0x50/0x1e0 Bytes 28-31 of 32 are uninitialized Memory access of size 32 starts at ffff88810098f570 ===================================================== Fixes: 83e2ec765be03e8a ("VMCI: doorbell implementation.") Cc: <stable@vger.kernel.org> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Link: https://lore.kernel.org/r/20210402121742.3917-1-penguin-kernel@I-love.SAKURA.ne.jp Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-04-02 12:17:41 +00:00
struct vmci_notify_bm_set_msg bitmap_set_msg = { };
bitmap_set_msg.hdr.dst = vmci_make_handle(VMCI_HYPERVISOR_CONTEXT_ID,
VMCI_SET_NOTIFY_BITMAP);
bitmap_set_msg.hdr.src = VMCI_ANON_SRC_HANDLE;
bitmap_set_msg.hdr.payload_size = sizeof(bitmap_set_msg) -
VMCI_DG_HEADERSIZE;
if (vmci_use_ppn64())
bitmap_set_msg.bitmap_ppn64 = bitmap_ppn;
else
bitmap_set_msg.bitmap_ppn32 = (u32) bitmap_ppn;
result = vmci_send_datagram(&bitmap_set_msg.hdr);
if (result != VMCI_SUCCESS) {
pr_devel("Failed to register (PPN=%llu) as notification bitmap (error=%d)\n",
bitmap_ppn, result);
return false;
}
return true;
}
/*
* Executes or schedules the handlers for a given notify index.
*/
static void dbell_fire_entries(u32 notify_idx)
{
u32 bucket = VMCI_DOORBELL_HASH(notify_idx);
struct dbell_entry *dbell;
spin_lock_bh(&vmci_doorbell_it.lock);
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
hlist_for_each_entry(dbell, &vmci_doorbell_it.entries[bucket], node) {
if (dbell->idx == notify_idx &&
atomic_read(&dbell->active) == 1) {
if (dbell->run_delayed) {
vmci_resource_get(&dbell->resource);
VMCI: Release resource if the work is already queued Francois reported that VMware balloon gets stuck after a balloon reset, when the VMCI doorbell is removed. A similar error can occur when the balloon driver is removed with the following splat: [ 1088.622000] INFO: task modprobe:3565 blocked for more than 120 seconds. [ 1088.622035] Tainted: G W 5.2.0 #4 [ 1088.622087] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1088.622205] modprobe D 0 3565 1450 0x00000000 [ 1088.622210] Call Trace: [ 1088.622246] __schedule+0x2a8/0x690 [ 1088.622248] schedule+0x2d/0x90 [ 1088.622250] schedule_timeout+0x1d3/0x2f0 [ 1088.622252] wait_for_completion+0xba/0x140 [ 1088.622320] ? wake_up_q+0x80/0x80 [ 1088.622370] vmci_resource_remove+0xb9/0xc0 [vmw_vmci] [ 1088.622373] vmci_doorbell_destroy+0x9e/0xd0 [vmw_vmci] [ 1088.622379] vmballoon_vmci_cleanup+0x6e/0xf0 [vmw_balloon] [ 1088.622381] vmballoon_exit+0x18/0xcc8 [vmw_balloon] [ 1088.622394] __x64_sys_delete_module+0x146/0x280 [ 1088.622408] do_syscall_64+0x5a/0x130 [ 1088.622410] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 1088.622415] RIP: 0033:0x7f54f62791b7 [ 1088.622421] Code: Bad RIP value. [ 1088.622421] RSP: 002b:00007fff2a949008 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0 [ 1088.622426] RAX: ffffffffffffffda RBX: 000055dff8b55d00 RCX: 00007f54f62791b7 [ 1088.622426] RDX: 0000000000000000 RSI: 0000000000000800 RDI: 000055dff8b55d68 [ 1088.622427] RBP: 000055dff8b55d00 R08: 00007fff2a947fb1 R09: 0000000000000000 [ 1088.622427] R10: 00007f54f62f5cc0 R11: 0000000000000206 R12: 000055dff8b55d68 [ 1088.622428] R13: 0000000000000001 R14: 000055dff8b55d68 R15: 00007fff2a94a3f0 The cause for the bug is that when the "delayed" doorbell is invoked, it takes a reference on the doorbell entry and schedules work that is supposed to run the appropriate code and drop the doorbell entry reference. The code ignores the fact that if the work is already queued, it will not be scheduled to run one more time. As a result one of the references would not be dropped. When the code waits for the reference to get to zero, during balloon reset or module removal, it gets stuck. Fix it. Drop the reference if schedule_work() indicates that the work is already queued. Note that this bug got more apparent (or apparent at all) due to commit ce664331b248 ("vmw_balloon: VMCI_DOORBELL_SET does not check status"). Fixes: 83e2ec765be03 ("VMCI: doorbell implementation.") Reported-by: Francois Rigault <rigault.francois@gmail.com> Cc: Jorgen Hansen <jhansen@vmware.com> Cc: Adit Ranadive <aditr@vmware.com> Cc: Alexios Zavras <alexios.zavras@intel.com> Cc: Vishnu DASA <vdasa@vmware.com> Cc: stable@vger.kernel.org Signed-off-by: Nadav Amit <namit@vmware.com> Reviewed-by: Vishnu Dasa <vdasa@vmware.com> Link: https://lore.kernel.org/r/20190820202638.49003-1-namit@vmware.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-08-20 20:26:38 +00:00
if (!schedule_work(&dbell->work))
vmci_resource_put(&dbell->resource);
} else {
dbell->notify_cb(dbell->client_data);
}
}
}
spin_unlock_bh(&vmci_doorbell_it.lock);
}
/*
* Scans the notification bitmap, collects pending notifications,
* resets the bitmap and invokes appropriate callbacks.
*/
void vmci_dbell_scan_notification_entries(u8 *bitmap)
{
u32 idx;
for (idx = 0; idx < max_notify_idx; idx++) {
if (bitmap[idx] & 0x1) {
bitmap[idx] &= ~1;
dbell_fire_entries(idx);
}
}
}
/*
* vmci_doorbell_create() - Creates a doorbell
* @handle: A handle used to track the resource. Can be invalid.
* @flags: Flag that determines context of callback.
* @priv_flags: Privileges flags.
* @notify_cb: The callback to be ivoked when the doorbell fires.
* @client_data: A parameter to be passed to the callback.
*
* Creates a doorbell with the given callback. If the handle is
* VMCI_INVALID_HANDLE, a free handle will be assigned, if
* possible. The callback can be run immediately (potentially with
* locks held - the default) or delayed (in a kernel thread) by
* specifying the flag VMCI_FLAG_DELAYED_CB. If delayed execution
* is selected, a given callback may not be run if the kernel is
* unable to allocate memory for the delayed execution (highly
* unlikely).
*/
int vmci_doorbell_create(struct vmci_handle *handle,
u32 flags,
u32 priv_flags,
vmci_callback notify_cb, void *client_data)
{
struct dbell_entry *entry;
struct vmci_handle new_handle;
int result;
if (!handle || !notify_cb || flags & ~VMCI_FLAG_DELAYED_CB ||
priv_flags & ~VMCI_PRIVILEGE_ALL_FLAGS)
return VMCI_ERROR_INVALID_ARGS;
entry = kmalloc(sizeof(*entry), GFP_KERNEL);
if (entry == NULL) {
pr_warn("Failed allocating memory for datagram entry\n");
return VMCI_ERROR_NO_MEM;
}
if (vmci_handle_is_invalid(*handle)) {
u32 context_id = vmci_get_context_id();
if (context_id == VMCI_INVALID_ID) {
pr_warn("Failed to get context ID\n");
result = VMCI_ERROR_NO_RESOURCES;
goto free_mem;
}
/* Let resource code allocate a free ID for us */
new_handle = vmci_make_handle(context_id, VMCI_INVALID_ID);
} else {
bool valid_context = false;
/*
* Validate the handle. We must do both of the checks below
* because we can be acting as both a host and a guest at the
* same time. We always allow the host context ID, since the
* host functionality is in practice always there with the
* unified driver.
*/
if (handle->context == VMCI_HOST_CONTEXT_ID ||
(vmci_guest_code_active() &&
vmci_get_context_id() == handle->context)) {
valid_context = true;
}
if (!valid_context || handle->resource == VMCI_INVALID_ID) {
pr_devel("Invalid argument (handle=0x%x:0x%x)\n",
handle->context, handle->resource);
result = VMCI_ERROR_INVALID_ARGS;
goto free_mem;
}
new_handle = *handle;
}
entry->idx = 0;
INIT_HLIST_NODE(&entry->node);
entry->priv_flags = priv_flags;
INIT_WORK(&entry->work, dbell_delayed_dispatch);
entry->run_delayed = flags & VMCI_FLAG_DELAYED_CB;
entry->notify_cb = notify_cb;
entry->client_data = client_data;
atomic_set(&entry->active, 0);
result = vmci_resource_add(&entry->resource,
VMCI_RESOURCE_TYPE_DOORBELL,
new_handle);
if (result != VMCI_SUCCESS) {
pr_warn("Failed to add new resource (handle=0x%x:0x%x), error: %d\n",
new_handle.context, new_handle.resource, result);
goto free_mem;
}
new_handle = vmci_resource_handle(&entry->resource);
if (vmci_guest_code_active()) {
dbell_index_table_add(entry);
result = dbell_link(new_handle, entry->idx);
if (VMCI_SUCCESS != result)
goto destroy_resource;
atomic_set(&entry->active, 1);
}
*handle = new_handle;
return result;
destroy_resource:
dbell_index_table_remove(entry);
vmci_resource_remove(&entry->resource);
free_mem:
kfree(entry);
return result;
}
EXPORT_SYMBOL_GPL(vmci_doorbell_create);
/*
* vmci_doorbell_destroy() - Destroy a doorbell.
* @handle: The handle tracking the resource.
*
* Destroys a doorbell previously created with vmcii_doorbell_create. This
* operation may block waiting for a callback to finish.
*/
int vmci_doorbell_destroy(struct vmci_handle handle)
{
struct dbell_entry *entry;
struct vmci_resource *resource;
if (vmci_handle_is_invalid(handle))
return VMCI_ERROR_INVALID_ARGS;
resource = vmci_resource_by_handle(handle,
VMCI_RESOURCE_TYPE_DOORBELL);
if (!resource) {
pr_devel("Failed to destroy doorbell (handle=0x%x:0x%x)\n",
handle.context, handle.resource);
return VMCI_ERROR_NOT_FOUND;
}
entry = container_of(resource, struct dbell_entry, resource);
if (!hlist_unhashed(&entry->node)) {
int result;
dbell_index_table_remove(entry);
result = dbell_unlink(handle);
if (VMCI_SUCCESS != result) {
/*
* The only reason this should fail would be
* an inconsistency between guest and
* hypervisor state, where the guest believes
* it has an active registration whereas the
* hypervisor doesn't. One case where this may
* happen is if a doorbell is unregistered
* following a hibernation at a time where the
* doorbell state hasn't been restored on the
* hypervisor side yet. Since the handle has
* now been removed in the guest, we just
* print a warning and return success.
*/
pr_devel("Unlink of doorbell (handle=0x%x:0x%x) unknown by hypervisor (error=%d)\n",
handle.context, handle.resource, result);
}
}
/*
* Now remove the resource from the table. It might still be in use
* after this, in a callback or still on the delayed work queue.
*/
vmci_resource_put(&entry->resource);
vmci_resource_remove(&entry->resource);
kfree(entry);
return VMCI_SUCCESS;
}
EXPORT_SYMBOL_GPL(vmci_doorbell_destroy);
/*
* vmci_doorbell_notify() - Ring the doorbell (and hide in the bushes).
* @dst: The handlle identifying the doorbell resource
* @priv_flags: Priviledge flags.
*
* Generates a notification on the doorbell identified by the
* handle. For host side generation of notifications, the caller
* can specify what the privilege of the calling side is.
*/
int vmci_doorbell_notify(struct vmci_handle dst, u32 priv_flags)
{
int retval;
enum vmci_route route;
struct vmci_handle src;
if (vmci_handle_is_invalid(dst) ||
(priv_flags & ~VMCI_PRIVILEGE_ALL_FLAGS))
return VMCI_ERROR_INVALID_ARGS;
src = VMCI_INVALID_HANDLE;
retval = vmci_route(&src, &dst, false, &route);
if (retval < VMCI_SUCCESS)
return retval;
if (VMCI_ROUTE_AS_HOST == route)
return vmci_ctx_notify_dbell(VMCI_HOST_CONTEXT_ID,
dst, priv_flags);
if (VMCI_ROUTE_AS_GUEST == route)
return dbell_notify_as_guest(dst, priv_flags);
pr_warn("Unknown route (%d) for doorbell\n", route);
return VMCI_ERROR_DST_UNREACHABLE;
}
EXPORT_SYMBOL_GPL(vmci_doorbell_notify);