IOMMU: RMRR (VT-d) and unity map (AMD-Vi) handling issues T[his CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] Certain PCI devices in a system might be assigned Reserved Memory Regions (specified via Reserved Memory Region Reporting, "RMRR") for Intel VT-d or Unity Mapping ranges for AMD-Vi. These are typically used for platform tasks such as legacy USB emulation. Since the precise purpose of these regions is unknown, once a device associated with such a region is active, the mappings of these regions need to remain continuouly accessible by the device. This requirement has been violated. Subsequent DMA or interrupts from the device may have unpredictable behaviour, ranging from IOMMU faults to memory corruption.

hw/usb/hcd-ohci.c in QEMU 5.0.0 has a stack-based buffer over-read via values obtained from the host controller driver.

The (1) sparc_mmap_check function in arch/sparc/kernel/sys_sparc.c and the (2) sparc64_mmap_check function in arch/sparc64/kernel/sys_sparc.c, in the Linux kernel 2.4 before 2.4.36.5 and 2.6 before 2.6.25.3, omit some virtual-address range (aka span) checks when the mmap MAP_FIXED bit is not set, which allows local users to cause a denial of service (panic) via unspecified mmap calls.

A TOCTOU mismatch in the NFS client code in the Linux kernel before 5.8.3 could be used by local attackers to corrupt memory or possibly have unspecified other impact because a size check is in fs/nfs/nfs4proc.c instead of fs/nfs/nfs4xdr.c, aka CID-b4487b935452.

The (1) memc_save_get_next_page, (2) tmemc_restore_put_page and (3) tmemc_restore_flush_page functions in the Transcendent Memory (TMEM) in Xen 4.0, 4.1, and 4.2 do not check for negative id pools, which allows local guest OS users to cause a denial of service (memory corruption and host crash) or possibly execute arbitrary code via unspecified vectors. NOTE: this issue was originally published as part of CVE-2012-3497, which was too general; CVE-2012-3497 has been SPLIT into this ID and others.

When using Apache Tomcat versions 10.0.0-M1 to 10.0.0-M4, 9.0.0.M1 to 9.0.34, 8.5.0 to 8.5.54 and 7.0.0 to 7.0.103 if a) an attacker is able to control the contents and name of a file on the server; and b) the server is configured to use the PersistenceManager with a FileStore; and c) the PersistenceManager is configured with sessionAttributeValueClassNameFilter="null" (the default unless a SecurityManager is used) or a sufficiently lax filter to allow the attacker provided object to be deserialized; and d) the attacker knows the relative file path from the storage location used by FileStore to the file the attacker has control over; then, using a specifically crafted request, the attacker will be able to trigger remote code execution via deserialization of the file under their control. Note that all of conditions a) to d) must be true for the attack to succeed.

The fix for bug CVE-2020-9484 introduced a time of check, time of use vulnerability into Apache Tomcat 10.1.0-M1 to 10.1.0-M8, 10.0.0-M5 to 10.0.14, 9.0.35 to 9.0.56 and 8.5.55 to 8.5.73 that allowed a local attacker to perform actions with the privileges of the user that the Tomcat process is using. This issue is only exploitable when Tomcat is configured to persist sessions using the FileStore.

The (1) tmemc_save_get_next_page and (2) tmemc_save_get_next_inv functions and the (3) TMEMC_SAVE_GET_POOL_UUID sub-operation in the Transcendent Memory (TMEM) in Xen 4.0, 4.1, and 4.2 "do not check incoming guest output buffer pointers," which allows local guest OS users to cause a denial of service (memory corruption and host crash) or execute arbitrary code via unspecified vectors. NOTE: this issue was originally published as part of CVE-2012-3497, which was too general; CVE-2012-3497 has been SPLIT into this ID and others.

A use-after-free vulnerability in the Linux kernel's af_unix component can be exploited to achieve local privilege escalation. The unix_stream_sendpage() function tries to add data to the last skb in the peer's recv queue without locking the queue. Thus there is a race where unix_stream_sendpage() could access an skb locklessly that is being released by garbage collection, resulting in use-after-free. We recommend upgrading past commit 790c2f9d15b594350ae9bca7b236f2b1859de02c.

The do_tmem_control function in the Transcendent Memory (TMEM) in Xen 4.0, 4.1, and 4.2 does not properly check privileges, which allows local guest OS users to access control stack operations via unspecified vectors. NOTE: this issue was originally published as part of CVE-2012-3497, which was too general; CVE-2012-3497 has been SPLIT into this ID and others.

Insufficient data validation in webUI in Google Chrome on ChromeOS prior to 86.0.4240.75 allowed a local attacker to bypass content security policy via a crafted HTML page.

GRUB2 fails to validate kernel signature when booted directly without shim, allowing secure boot to be bypassed. This only affects systems where the kernel signing certificate has been imported directly into the secure boot database and the GRUB image is booted directly without the use of shim. This issue affects GRUB2 version 2.04 and prior versions.

Blueman is a GTK+ Bluetooth Manager. In Blueman before 2.1.4, the DhcpClient method of the D-Bus interface to blueman-mechanism is prone to an argument injection vulnerability. The impact highly depends on the system configuration. If Polkit-1 is disabled and for versions lower than 2.0.6, any local user can possibly exploit this. If Polkit-1 is enabled for version 2.0.6 and later, a possible attacker needs to be allowed to use the `org.blueman.dhcp.client` action. That is limited to users in the wheel group in the shipped rules file that do have the privileges anyway. On systems with ISC DHCP client (dhclient), attackers can pass arguments to `ip link` with the interface name that can e.g. be used to bring down an interface or add an arbitrary XDP/BPF program. On systems with dhcpcd and without ISC DHCP client, attackers can even run arbitrary scripts by passing `-c/path/to/script` as an interface name. Patches are included in 2.1.4 and master that change the DhcpClient D-Bus method(s) to accept BlueZ network object paths instead of network interface names. A backport to 2.0(.8) is also available. As a workaround, make sure that Polkit-1-support is enabled and limit privileges for the `org.blueman.dhcp.client` action to users that are able to run arbitrary commands as root anyway in /usr/share/polkit-1/rules.d/blueman.rules.

hw/net/xgmac.c in the XGMAC Ethernet controller in QEMU before 07-20-2020 has a buffer overflow. This occurs during packet transmission and affects the highbank and midway emulated machines. A guest user or process could use this flaw to crash the QEMU process on the host, resulting in a denial of service or potential privileged code execution. This was fixed in commit 5519724a13664b43e225ca05351c60b4468e4555.

A use-after-free vulnerability in the Linux kernel's netfilter: nf_tables component can be exploited to achieve local privilege escalation. Due to a race condition between nf_tables netlink control plane transaction and nft_set element garbage collection, it is possible to underflow the reference counter causing a use-after-free vulnerability. We recommend upgrading past commit 3e91b0ebd994635df2346353322ac51ce84ce6d8.

In emulation_proc_handler of armv8_deprecated.c, there is a possible way to corrupt memory due to a race condition. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.Product: AndroidVersions: Android kernelAndroid ID: A-237540956References: Upstream kernel

The default configuration of the apache2 package in Debian GNU/Linux squeeze before 2.2.16-6+squeeze7, wheezy before 2.2.22-4, and sid before 2.2.22-4, when mod_php or mod_rivet is used, provides example scripts under the doc/ URI, which might allow local users to conduct cross-site scripting (XSS) attacks, gain privileges, or obtain sensitive information via vectors involving localhost HTTP requests to the Apache HTTP Server.

It was found that some PostgreSQL extensions did not use search_path safely in their installation script. An attacker with sufficient privileges could use this flaw to trick an administrator into executing a specially crafted script, during the installation or update of such extension. This affects PostgreSQL versions before 12.4, before 11.9, before 10.14, before 9.6.19, and before 9.5.23.

ext/fts3/fts3.c in SQLite before 3.32.0 has a use-after-free in fts3EvalNextRow, related to the snippet feature.

A flaw in Linux Kernel found in nfcmrvl_nci_unregister_dev() in drivers/nfc/nfcmrvl/main.c can lead to use after free both read or write when non synchronized between cleanup routine and firmware download routine.

There is a use-after-free in kernel versions before 5.5 due to a race condition between the release of ptp_clock and cdev while resource deallocation. When a (high privileged) process allocates a ptp device file (like /dev/ptpX) and voluntarily goes to sleep. During this time if the underlying device is removed, it can cause an exploitable condition as the process wakes up to terminate and clean all attached files. The system crashes due to the cdev structure being invalid (as already freed) which is pointed to by the inode.

In the Linux kernel, pick_next_rt_entity() may return a type confused entry, not detected by the BUG_ON condition, as the confused entry will not be NULL, but list_head.The buggy error condition would lead to a type confused entry with the list head,which would then be used as a type confused sched_rt_entity,causing memory corruption.

Flatpak before 1.0.7, and 1.1.x and 1.2.x before 1.2.3, exposes /proc in the apply_extra script sandbox, which allows attackers to modify a host-side executable file.

open-vm-tools contains a file descriptor hijack vulnerability in the vmware-user-suid-wrapper. A malicious actor with non-root privileges may be able to hijack the /dev/uinput file descriptor allowing them to simulate user inputs.

sysstat is a set of system performance tools for the Linux operating system. On 32 bit systems, in versions 9.1.16 and newer but prior to 12.7.1, allocate_structures contains a size_t overflow in sa_common.c. The allocate_structures function insufficiently checks bounds before arithmetic multiplication, allowing for an overflow in the size allocated for the buffer representing system activities. This issue may lead to Remote Code Execution (RCE). This issue has been patched in version 12.7.1.

A flaw was found in the way the dumpable flag setting was handled when certain SUID binaries executed its descendants. The prerequisite is a SUID binary that sets real UID equal to effective UID, and real GID equal to effective GID. The descendant will then have a dumpable value set to 1. As a result, if the descendant process crashes and core_pattern is set to a relative value, its core dump is stored in the current directory with uid:gid permissions. An unprivileged local user with eligible root SUID binary could use this flaw to place core dumps into root-owned directories, potentially resulting in escalation of privileges.

The npm package "tar" (aka node-tar) before versions 4.4.16, 5.0.8, and 6.1.7 has an arbitrary file creation/overwrite and arbitrary code execution vulnerability. node-tar aims to guarantee that any file whose location would be modified by a symbolic link is not extracted. This is, in part, achieved by ensuring that extracted directories are not symlinks. Additionally, in order to prevent unnecessary stat calls to determine whether a given path is a directory, paths are cached when directories are created. This logic was insufficient when extracting tar files that contained both a directory and a symlink with the same name as the directory, where the symlink and directory names in the archive entry used backslashes as a path separator on posix systems. The cache checking logic used both `\` and `/` characters as path separators, however `\` is a valid filename character on posix systems. By first creating a directory, and then replacing that directory with a symlink, it was thus possible to bypass node-tar symlink checks on directories, essentially allowing an untrusted tar file to symlink into an arbitrary location and subsequently extracting arbitrary files into that location, thus allowing arbitrary file creation and overwrite. Additionally, a similar confusion could arise on case-insensitive filesystems. If a tar archive contained a directory at `FOO`, followed by a symbolic link named `foo`, then on case-insensitive file systems, the creation of the symbolic link would remove the directory from the filesystem, but _not_ from the internal directory cache, as it would not be treated as a cache hit. A subsequent file entry within the `FOO` directory would then be placed in the target of the symbolic link, thinking that the directory had already been created. These issues were addressed in releases 4.4.16, 5.0.8 and 6.1.7. The v3 branch of node-tar has been deprecated and did not receive patches for these issues. If you are still using a v3 release we recommend you update to a more recent version of node-tar. If this is not possible, a workaround is available in the referenced GHSA-9r2w-394v-53qc.

Firejail before 0.9.64.4 allows attackers to bypass intended access restrictions because there is a TOCTOU race condition between a stat operation and an OverlayFS mount operation.

The fix for CVE-2020-9484 was incomplete. When using Apache Tomcat 10.0.0-M1 to 10.0.0, 9.0.0.M1 to 9.0.41, 8.5.0 to 8.5.61 or 7.0.0. to 7.0.107 with a configuration edge case that was highly unlikely to be used, the Tomcat instance was still vulnerable to CVE-2020-9494. Note that both the previously published prerequisites for CVE-2020-9484 and the previously published mitigations for CVE-2020-9484 also apply to this issue.

The OPVPWrapper::loadDriver function in oprs/OPVPWrapper.cxx in the pdftoopvp filter in CUPS and cups-filters before 1.0.47 allows local users to gain privileges via a Trojan horse driver in the same directory as the PDF file.

Race condition in the rmtree and remove_tree functions in the File-Path module before 2.13 for Perl allows attackers to set the mode on arbitrary files via vectors involving directory-permission loosening logic.

A race condition existed in the snapd 2.54.2 snap-confine binary when preparing a private mount namespace for a snap. This could allow a local attacker to gain root privileges by bind-mounting their own contents inside the snap's private mount namespace and causing snap-confine to execute arbitrary code and hence gain privilege escalation. Fixed in snapd versions 2.54.3+18.04, 2.54.3+20.04 and 2.54.3+21.10.1

The simplified implementation of blocking reads and writes introduced in Tomcat 10 and back-ported to Tomcat 9.0.47 onwards exposed a long standing (but extremely hard to trigger) concurrency bug in Apache Tomcat 10.1.0 to 10.1.0-M12, 10.0.0-M1 to 10.0.18, 9.0.0-M1 to 9.0.60 and 8.5.0 to 8.5.77 that could cause client connections to share an Http11Processor instance resulting in responses, or part responses, to be received by the wrong client.

By misusing a race in our notification code, an attacker could have forcefully hidden the notification for pages that had received full screen and pointer lock access, which could have been used for spoofing attacks. This vulnerability affects Thunderbird < 91.4.0, Firefox ESR < 91.4.0, and Firefox < 95.

Incorrect security UI in popup blocker in Google Chrome on iOS prior to 75.0.3770.80 allowed a remote attacker to bypass navigation restrictions via a crafted HTML page.

A flaw was found in the QXL display device emulation in QEMU. A double fetch of guest controlled values `cursor->header.width` and `cursor->header.height` can lead to the allocation of a small cursor object followed by a subsequent heap-based buffer overflow. A malicious privileged guest user could use this flaw to crash the QEMU process on the host or potentially execute arbitrary code within the context of the QEMU process.

A race condition was found in util-linux before 2.32.1 in the way su handled the management of child processes. A local authenticated attacker could use this flaw to kill other processes with root privileges under specific conditions.

systemd does not properly use D-Bus for communication with a polkit authority, which allows local users to bypass intended access restrictions by leveraging a PolkitUnixProcess PolkitSubject race condition via a (1) setuid process or (2) pkexec process, a related issue to CVE-2013-4288.

An issue was discovered in Xen through 4.9.x allowing x86 PV guest OS users to execute arbitrary code on the host OS because of a race condition that can cause a stale TLB entry.

A domain cleanup issue was discovered in the C xenstore daemon (aka cxenstored) in Xen through 4.9.x. When shutting down a VM with a stubdomain, a race in cxenstored may cause a double-free. The xenstored daemon may crash, resulting in a DoS of any parts of the system relying on it (including domain creation / destruction, ballooning, device changes, etc.).

A use-after-free flaw was found in the Linux kernel’s Bluetooth subsystem in the way user calls connect to the socket and disconnect simultaneously due to a race condition. This flaw allows a user to crash the system or escalate their privileges. The highest threat from this vulnerability is to confidentiality, integrity, as well as system availability.

Go before 1.15.15 and 1.16.x before 1.16.7 has a race condition that can lead to a net/http/httputil ReverseProxy panic upon an ErrAbortHandler abort.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nf_tables: Fix potential data-race in __nft_flowtable_type_get() nft_unregister_flowtable_type() within nf_flow_inet_module_exit() can concurrent with __nft_flowtable_type_get() within nf_tables_newflowtable(). And thhere is not any protection when iterate over nf_tables_flowtables list in __nft_flowtable_type_get(). Therefore, there is pertential data-race of nf_tables_flowtables list entry. Use list_for_each_entry_rcu() to iterate over nf_tables_flowtables list in __nft_flowtable_type_get(), and use rcu_read_lock() in the caller nft_flowtable_type_get() to protect the entire type query process.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nf_tables: flush pending destroy work before exit_net release Similar to 2c9f0293280e ("netfilter: nf_tables: flush pending destroy work before netlink notifier") to address a race between exit_net and the destroy workqueue. The trace below shows an element to be released via destroy workqueue while exit_net path (triggered via module removal) has already released the set that is used in such transaction. [ 1360.547789] BUG: KASAN: slab-use-after-free in nf_tables_trans_destroy_work+0x3f5/0x590 [nf_tables] [ 1360.547861] Read of size 8 at addr ffff888140500cc0 by task kworker/4:1/152465 [ 1360.547870] CPU: 4 PID: 152465 Comm: kworker/4:1 Not tainted 6.8.0+ #359 [ 1360.547882] Workqueue: events nf_tables_trans_destroy_work [nf_tables] [ 1360.547984] Call Trace: [ 1360.547991] <TASK> [ 1360.547998] dump_stack_lvl+0x53/0x70 [ 1360.548014] print_report+0xc4/0x610 [ 1360.548026] ? __virt_addr_valid+0xba/0x160 [ 1360.548040] ? __pfx__raw_spin_lock_irqsave+0x10/0x10 [ 1360.548054] ? nf_tables_trans_destroy_work+0x3f5/0x590 [nf_tables] [ 1360.548176] kasan_report+0xae/0xe0 [ 1360.548189] ? nf_tables_trans_destroy_work+0x3f5/0x590 [nf_tables] [ 1360.548312] nf_tables_trans_destroy_work+0x3f5/0x590 [nf_tables] [ 1360.548447] ? __pfx_nf_tables_trans_destroy_work+0x10/0x10 [nf_tables] [ 1360.548577] ? _raw_spin_unlock_irq+0x18/0x30 [ 1360.548591] process_one_work+0x2f1/0x670 [ 1360.548610] worker_thread+0x4d3/0x760 [ 1360.548627] ? __pfx_worker_thread+0x10/0x10 [ 1360.548640] kthread+0x16b/0x1b0 [ 1360.548653] ? __pfx_kthread+0x10/0x10 [ 1360.548665] ret_from_fork+0x2f/0x50 [ 1360.548679] ? __pfx_kthread+0x10/0x10 [ 1360.548690] ret_from_fork_asm+0x1a/0x30 [ 1360.548707] </TASK> [ 1360.548719] Allocated by task 192061: [ 1360.548726] kasan_save_stack+0x20/0x40 [ 1360.548739] kasan_save_track+0x14/0x30 [ 1360.548750] __kasan_kmalloc+0x8f/0xa0 [ 1360.548760] __kmalloc_node+0x1f1/0x450 [ 1360.548771] nf_tables_newset+0x10c7/0x1b50 [nf_tables] [ 1360.548883] nfnetlink_rcv_batch+0xbc4/0xdc0 [nfnetlink] [ 1360.548909] nfnetlink_rcv+0x1a8/0x1e0 [nfnetlink] [ 1360.548927] netlink_unicast+0x367/0x4f0 [ 1360.548935] netlink_sendmsg+0x34b/0x610 [ 1360.548944] ____sys_sendmsg+0x4d4/0x510 [ 1360.548953] ___sys_sendmsg+0xc9/0x120 [ 1360.548961] __sys_sendmsg+0xbe/0x140 [ 1360.548971] do_syscall_64+0x55/0x120 [ 1360.548982] entry_SYSCALL_64_after_hwframe+0x55/0x5d [ 1360.548994] Freed by task 192222: [ 1360.548999] kasan_save_stack+0x20/0x40 [ 1360.549009] kasan_save_track+0x14/0x30 [ 1360.549019] kasan_save_free_info+0x3b/0x60 [ 1360.549028] poison_slab_object+0x100/0x180 [ 1360.549036] __kasan_slab_free+0x14/0x30 [ 1360.549042] kfree+0xb6/0x260 [ 1360.549049] __nft_release_table+0x473/0x6a0 [nf_tables] [ 1360.549131] nf_tables_exit_net+0x170/0x240 [nf_tables] [ 1360.549221] ops_exit_list+0x50/0xa0 [ 1360.549229] free_exit_list+0x101/0x140 [ 1360.549236] unregister_pernet_operations+0x107/0x160 [ 1360.549245] unregister_pernet_subsys+0x1c/0x30 [ 1360.549254] nf_tables_module_exit+0x43/0x80 [nf_tables] [ 1360.549345] __do_sys_delete_module+0x253/0x370 [ 1360.549352] do_syscall_64+0x55/0x120 [ 1360.549360] entry_SYSCALL_64_after_hwframe+0x55/0x5d (gdb) list *__nft_release_table+0x473 0x1e033 is in __nft_release_table (net/netfilter/nf_tables_api.c:11354). 11349 list_for_each_entry_safe(flowtable, nf, &table->flowtables, list) { 11350 list_del(&flowtable->list); 11351 nft_use_dec(&table->use); 11352 nf_tables_flowtable_destroy(flowtable); 11353 } 11354 list_for_each_entry_safe(set, ns, &table->sets, list) { 11355 list_del(&set->list); 11356 nft_use_dec(&table->use); 11357 if (set->flags & (NFT_SET_MAP | NFT_SET_OBJECT)) 11358 nft_map_deactivat ---truncated---

An issue was discovered in the Linux kernel before 5.11.7. usbip_sockfd_store in drivers/usb/usbip/stub_dev.c allows attackers to cause a denial of service (GPF) because the stub-up sequence has race conditions during an update of the local and shared status, aka CID-9380afd6df70.

A race condition was discovered in get_old_root in fs/btrfs/ctree.c in the Linux kernel through 5.11.8. It allows attackers to cause a denial of service (BUG) because of a lack of locking on an extent buffer before a cloning operation, aka CID-dbcc7d57bffc.

Race condition in the sandbox launcher implementation in Google Chrome before 11.0.696.57 on Linux allows remote attackers to cause a denial of service or possibly have unspecified other impact via unknown vectors.

In the Linux kernel, the following vulnerability has been resolved: wireguard: receive: annotate data-race around receiving_counter.counter Syzkaller with KCSAN identified a data-race issue when accessing keypair->receiving_counter.counter. Use READ_ONCE() and WRITE_ONCE() annotations to mark the data race as intentional. BUG: KCSAN: data-race in wg_packet_decrypt_worker / wg_packet_rx_poll write to 0xffff888107765888 of 8 bytes by interrupt on cpu 0: counter_validate drivers/net/wireguard/receive.c:321 [inline] wg_packet_rx_poll+0x3ac/0xf00 drivers/net/wireguard/receive.c:461 __napi_poll+0x60/0x3b0 net/core/dev.c:6536 napi_poll net/core/dev.c:6605 [inline] net_rx_action+0x32b/0x750 net/core/dev.c:6738 __do_softirq+0xc4/0x279 kernel/softirq.c:553 do_softirq+0x5e/0x90 kernel/softirq.c:454 __local_bh_enable_ip+0x64/0x70 kernel/softirq.c:381 __raw_spin_unlock_bh include/linux/spinlock_api_smp.h:167 [inline] _raw_spin_unlock_bh+0x36/0x40 kernel/locking/spinlock.c:210 spin_unlock_bh include/linux/spinlock.h:396 [inline] ptr_ring_consume_bh include/linux/ptr_ring.h:367 [inline] wg_packet_decrypt_worker+0x6c5/0x700 drivers/net/wireguard/receive.c:499 process_one_work kernel/workqueue.c:2633 [inline] ... read to 0xffff888107765888 of 8 bytes by task 3196 on cpu 1: decrypt_packet drivers/net/wireguard/receive.c:252 [inline] wg_packet_decrypt_worker+0x220/0x700 drivers/net/wireguard/receive.c:501 process_one_work kernel/workqueue.c:2633 [inline] process_scheduled_works+0x5b8/0xa30 kernel/workqueue.c:2706 worker_thread+0x525/0x730 kernel/workqueue.c:2787 ...

In the Linux kernel, the following vulnerability has been resolved: hv_netvsc: Fix race condition between netvsc_probe and netvsc_remove In commit ac5047671758 ("hv_netvsc: Disable NAPI before closing the VMBus channel"), napi_disable was getting called for all channels, including all subchannels without confirming if they are enabled or not. This caused hv_netvsc getting hung at napi_disable, when netvsc_probe() has finished running but nvdev->subchan_work has not started yet. netvsc_subchan_work() -> rndis_set_subchannel() has not created the sub-channels and because of that netvsc_sc_open() is not running. netvsc_remove() calls cancel_work_sync(&nvdev->subchan_work), for which netvsc_subchan_work did not run. netif_napi_add() sets the bit NAPI_STATE_SCHED because it ensures NAPI cannot be scheduled. Then netvsc_sc_open() -> napi_enable will clear the NAPIF_STATE_SCHED bit, so it can be scheduled. napi_disable() does the opposite. Now during netvsc_device_remove(), when napi_disable is called for those subchannels, napi_disable gets stuck on infinite msleep. This fix addresses this problem by ensuring that napi_disable() is not getting called for non-enabled NAPI struct. But netif_napi_del() is still necessary for these non-enabled NAPI struct for cleanup purpose. Call trace: [ 654.559417] task:modprobe state:D stack: 0 pid: 2321 ppid: 1091 flags:0x00004002 [ 654.568030] Call Trace: [ 654.571221] <TASK> [ 654.573790] __schedule+0x2d6/0x960 [ 654.577733] schedule+0x69/0xf0 [ 654.581214] schedule_timeout+0x87/0x140 [ 654.585463] ? __bpf_trace_tick_stop+0x20/0x20 [ 654.590291] msleep+0x2d/0x40 [ 654.593625] napi_disable+0x2b/0x80 [ 654.597437] netvsc_device_remove+0x8a/0x1f0 [hv_netvsc] [ 654.603935] rndis_filter_device_remove+0x194/0x1c0 [hv_netvsc] [ 654.611101] ? do_wait_intr+0xb0/0xb0 [ 654.615753] netvsc_remove+0x7c/0x120 [hv_netvsc] [ 654.621675] vmbus_remove+0x27/0x40 [hv_vmbus]

In the Linux kernel, the following vulnerability has been resolved: mm: swap: fix race between free_swap_and_cache() and swapoff() There was previously a theoretical window where swapoff() could run and teardown a swap_info_struct while a call to free_swap_and_cache() was running in another thread. This could cause, amongst other bad possibilities, swap_page_trans_huge_swapped() (called by free_swap_and_cache()) to access the freed memory for swap_map. This is a theoretical problem and I haven't been able to provoke it from a test case. But there has been agreement based on code review that this is possible (see link below). Fix it by using get_swap_device()/put_swap_device(), which will stall swapoff(). There was an extra check in _swap_info_get() to confirm that the swap entry was not free. This isn't present in get_swap_device() because it doesn't make sense in general due to the race between getting the reference and swapoff. So I've added an equivalent check directly in free_swap_and_cache(). Details of how to provoke one possible issue (thanks to David Hildenbrand for deriving this): --8<----- __swap_entry_free() might be the last user and result in "count == SWAP_HAS_CACHE". swapoff->try_to_unuse() will stop as soon as soon as si->inuse_pages==0. So the question is: could someone reclaim the folio and turn si->inuse_pages==0, before we completed swap_page_trans_huge_swapped(). Imagine the following: 2 MiB folio in the swapcache. Only 2 subpages are still references by swap entries. Process 1 still references subpage 0 via swap entry. Process 2 still references subpage 1 via swap entry. Process 1 quits. Calls free_swap_and_cache(). -> count == SWAP_HAS_CACHE [then, preempted in the hypervisor etc.] Process 2 quits. Calls free_swap_and_cache(). -> count == SWAP_HAS_CACHE Process 2 goes ahead, passes swap_page_trans_huge_swapped(), and calls __try_to_reclaim_swap(). __try_to_reclaim_swap()->folio_free_swap()->delete_from_swap_cache()-> put_swap_folio()->free_swap_slot()->swapcache_free_entries()-> swap_entry_free()->swap_range_free()-> ... WRITE_ONCE(si->inuse_pages, si->inuse_pages - nr_entries); What stops swapoff to succeed after process 2 reclaimed the swap cache but before process1 finished its call to swap_page_trans_huge_swapped()? --8<-----