[ Upstream commit 0afec5e9cea732cb47014655685a2a47fb180c31 ]
When registering a new DMA MR after selecting the best aligned page size
for it, we iterate over the given sglist to split each entry to smaller,
aligned to the selected page size, DMA blocks.
In given circumstances where the sg entry and page size fit certain
sizes and the sg entry is not aligned to the selected page size, the
total size of the aligned pages we need to cover the sg entry is >= 4GB.
Under this circumstances, while iterating page aligned blocks, the
counter responsible for counting how much we advanced from the start of
the sg entry is overflowed because its type is u32 and we pass 4GB in
size. This can lead to an infinite loop inside the iterator function
because the overflow prevents the counter to be larger
than the size of the sg entry.
Fix the presented problem by changing the advancement condition to
eliminate overflow.
Backtrace:
[ 192.374329] efa_reg_user_mr_dmabuf
[ 192.376783] efa_register_mr
[ 192.382579] pgsz_bitmap 0xfffff000 rounddown 0x80000000
[ 192.386423] pg_sz [0x80000000] umem_length[0xc0000000]
[ 192.392657] start 0x0 length 0xc0000000 params.page_shift 31 params.page_num 3
[ 192.399559] hp_cnt[3], pages_in_hp[524288]
[ 192.403690] umem->sgt_append.sgt.nents[1]
[ 192.407905] number entries: [1], pg_bit: [31]
[ 192.411397] biter->__sg_nents [1] biter->__sg [0000000008b0c5d8]
[ 192.415601] biter->__sg_advance [665837568] sg_dma_len[3221225472]
[ 192.419823] biter->__sg_nents [1] biter->__sg [0000000008b0c5d8]
[ 192.423976] biter->__sg_advance [2813321216] sg_dma_len[3221225472]
[ 192.428243] biter->__sg_nents [1] biter->__sg [0000000008b0c5d8]
[ 192.432397] biter->__sg_advance [665837568] sg_dma_len[3221225472]
Fixes: a808273a49 ("RDMA/verbs: Add a DMA iterator to return aligned contiguous memory blocks")
Signed-off-by: Yonatan Nachum <ynachum@amazon.com>
Link: https://lore.kernel.org/r/20230109133711.13678-1-ynachum@amazon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit ecacb3751f254572af0009b9501e2cdc83a30b6a ]
When filling a cm_id entry, return "-EAGAIN" instead of 0 if the cm_id
doesn'the have the same port as requested, otherwise an incomplete entry
may be returned, which causes "rdam res show cm_id" to return an error.
For example on a machine with two rdma devices with "rping -C 1 -v -s"
running background, the "rdma" command fails:
$ rdma -V
rdma utility, iproute2-5.19.0
$ rdma res show cm_id
link mlx5_0/- cm-idn 0 state LISTEN ps TCP pid 28056 comm rping src-addr 0.0.0.0:7174
error: Protocol not available
While with this fix it succeeds:
$ rdma res show cm_id
link mlx5_0/- cm-idn 0 state LISTEN ps TCP pid 26395 comm rping src-addr 0.0.0.0:7174
link mlx5_1/- cm-idn 0 state LISTEN ps TCP pid 26395 comm rping src-addr 0.0.0.0:7174
Fixes: 00313983cd ("RDMA/nldev: provide detailed CM_ID information")
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Link: https://lore.kernel.org/r/a08e898cdac5e28428eb749a99d9d981571b8ea7.1667810736.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit eb83f502adb036cd56c27e13b9ca3b2aabfa790b ]
Commit 27cfde795a96 ("RDMA/cma: Fix arguments order in net device
validation") swapped the src and dst addresses in the call to
validate_net_dev().
As a consequence, the test in validate_ipv4_net_dev() to see if the
net_dev is the right one, is incorrect for port 1 <-> 2 communication when
the ports are on the same sub-net. This is fixed by denoting the
flowi4_oif as the device instead of the incoming one.
The bug has not been observed using IPv6 addresses.
Fixes: 27cfde795a96 ("RDMA/cma: Fix arguments order in net device validation")
Signed-off-by: Håkon Bugge <haakon.bugge@oracle.com>
Link: https://lore.kernel.org/r/20221012141542.16925-1-haakon.bugge@oracle.com
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 241f9a27e0fc0eaf23e3d52c8450f10648cd11f1 ]
Set 'iova' and 'length' on ib_mr in ib_uverbs and ib_core layers to let all
drivers have the members filled. Also, this commit removes redundancy in
the respective drivers.
Previously, commit 04c0a5fcfc ("IB/uverbs: Set IOVA on IB MR in uverbs
layer") changed to set 'iova', but seems to have missed 'length' and the
ib_core layer at that time.
Fixes: 04c0a5fcfc ("IB/uverbs: Set IOVA on IB MR in uverbs layer")
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://lore.kernel.org/r/20220921080844.1616883-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b7d95040c13f61a4a6a859c5355faf583eff9658 ]
The responder should always use WC's SLID as the dlid, to follow the
IB SPEC section "13.5.4.2 COMMON RESPONSE ACTIONS":
A responder always takes the following actions in constructing a
response packet:
- The SLID of the received packet is used as the DLID in the response
packet.
Fixes: ac3a949fb2 ("IB/CM: Set appropriate slid and dlid when handling CM request")
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Link: https://lore.kernel.org/r/cd17c240231e059d2fc07c17dfe555d548b917eb.1662631201.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 85eaeb5058f0f04dffb124c97c86b4f18db0b833 ]
Fix a nested dead lock as part of ODP flow by using mmput_async().
From the below call trace [1] can see that calling mmput() once we have
the umem_odp->umem_mutex locked as required by
ib_umem_odp_map_dma_and_lock() might trigger in the same task the
exit_mmap()->__mmu_notifier_release()->mlx5_ib_invalidate_range() which
may dead lock when trying to lock the same mutex.
Moving to use mmput_async() will solve the problem as the above
exit_mmap() flow will be called in other task and will be executed once
the lock will be available.
[1]
[64843.077665] task:kworker/u133:2 state:D stack: 0 pid:80906 ppid:
2 flags:0x00004000
[64843.077672] Workqueue: mlx5_ib_page_fault mlx5_ib_eqe_pf_action [mlx5_ib]
[64843.077719] Call Trace:
[64843.077722] <TASK>
[64843.077724] __schedule+0x23d/0x590
[64843.077729] schedule+0x4e/0xb0
[64843.077735] schedule_preempt_disabled+0xe/0x10
[64843.077740] __mutex_lock.constprop.0+0x263/0x490
[64843.077747] __mutex_lock_slowpath+0x13/0x20
[64843.077752] mutex_lock+0x34/0x40
[64843.077758] mlx5_ib_invalidate_range+0x48/0x270 [mlx5_ib]
[64843.077808] __mmu_notifier_release+0x1a4/0x200
[64843.077816] exit_mmap+0x1bc/0x200
[64843.077822] ? walk_page_range+0x9c/0x120
[64843.077828] ? __cond_resched+0x1a/0x50
[64843.077833] ? mutex_lock+0x13/0x40
[64843.077839] ? uprobe_clear_state+0xac/0x120
[64843.077860] mmput+0x5f/0x140
[64843.077867] ib_umem_odp_map_dma_and_lock+0x21b/0x580 [ib_core]
[64843.077931] pagefault_real_mr+0x9a/0x140 [mlx5_ib]
[64843.077962] pagefault_mr+0xb4/0x550 [mlx5_ib]
[64843.077992] pagefault_single_data_segment.constprop.0+0x2ac/0x560
[mlx5_ib]
[64843.078022] mlx5_ib_eqe_pf_action+0x528/0x780 [mlx5_ib]
[64843.078051] process_one_work+0x22b/0x3d0
[64843.078059] worker_thread+0x53/0x410
[64843.078065] ? process_one_work+0x3d0/0x3d0
[64843.078073] kthread+0x12a/0x150
[64843.078079] ? set_kthread_struct+0x50/0x50
[64843.078085] ret_from_fork+0x22/0x30
[64843.078093] </TASK>
Fixes: 36f30e486d ("IB/core: Improve ODP to use hmm_range_fault()")
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Link: https://lore.kernel.org/r/74d93541ea533ef7daec6f126deb1072500aeb16.1661251841.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 27cfde795a96aef1e859a5480489944b95421e46 ]
Fix the order of source and destination addresses when resolving the
route between server and client to validate use of correct net device.
The reverse order we had so far didn't actually validate the net device
as the server would try to resolve the route to itself, thus always
getting the server's net device.
The issue was discovered when running cm applications on a single host
between 2 interfaces with same subnet and source based routing rules.
When resolving the reverse route the source based route rules were
ignored.
Fixes: f887f2ac87 ("IB/cma: Validate routing of incoming requests")
Link: https://lore.kernel.org/r/1c1ec2277a131d277ebcceec987fd338d35b775f.1661251872.git.leonro@nvidia.com
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 7c4a539ec38f4ce400a0f3fcb5ff6c940fcd67bb ]
ib_destroy_qp() would called by ib_create_qp_user() if error, the former
contains ib_qp_usecnt_dec(), but ib_qp_usecnt_inc() was not called before.
So move ib_qp_usecnt_inc() into create_qp().
Fixes: d2b10794fc ("RDMA/core: Create clean QP creations interface for uverbs")
Link: https://lore.kernel.org/r/20220303024232.2847388-1-yajun.deng@linux.dev
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 22e9f71072fa605cbf033158db58e0790101928d upstream.
If the state is not idle then resolve_prepare_src() should immediately
fail and no change to global state should happen. However, it
unconditionally overwrites the src_addr trying to build a temporary any
address.
For instance if the state is already RDMA_CM_LISTEN then this will corrupt
the src_addr and would cause the test in cma_cancel_operation():
if (cma_any_addr(cma_src_addr(id_priv)) && !id_priv->cma_dev)
Which would manifest as this trace from syzkaller:
BUG: KASAN: use-after-free in __list_add_valid+0x93/0xa0 lib/list_debug.c:26
Read of size 8 at addr ffff8881546491e0 by task syz-executor.1/32204
CPU: 1 PID: 32204 Comm: syz-executor.1 Not tainted 5.12.0-rc8-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:79 [inline]
dump_stack+0x141/0x1d7 lib/dump_stack.c:120
print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:232
__kasan_report mm/kasan/report.c:399 [inline]
kasan_report.cold+0x7c/0xd8 mm/kasan/report.c:416
__list_add_valid+0x93/0xa0 lib/list_debug.c:26
__list_add include/linux/list.h:67 [inline]
list_add_tail include/linux/list.h:100 [inline]
cma_listen_on_all drivers/infiniband/core/cma.c:2557 [inline]
rdma_listen+0x787/0xe00 drivers/infiniband/core/cma.c:3751
ucma_listen+0x16a/0x210 drivers/infiniband/core/ucma.c:1102
ucma_write+0x259/0x350 drivers/infiniband/core/ucma.c:1732
vfs_write+0x28e/0xa30 fs/read_write.c:603
ksys_write+0x1ee/0x250 fs/read_write.c:658
do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
entry_SYSCALL_64_after_hwframe+0x44/0xae
This is indicating that an rdma_id_private was destroyed without doing
cma_cancel_listens().
Instead of trying to re-use the src_addr memory to indirectly create an
any address derived from the dst build one explicitly on the stack and
bind to that as any other normal flow would do. rdma_bind_addr() will copy
it over the src_addr once it knows the state is valid.
This is similar to commit bc0bdc5afa ("RDMA/cma: Do not change
route.addr.src_addr.ss_family")
Link: https://lore.kernel.org/r/0-v2-e975c8fd9ef2+11e-syz_cma_srcaddr_jgg@nvidia.com
Cc: stable@vger.kernel.org
Fixes: 732d41c545 ("RDMA/cma: Make the locking for automatic state transition more clear")
Reported-by: syzbot+c94a3675a626f6333d74@syzkaller.appspotmail.com
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 36e8169ec973359f671f9ec7213547059cae972e upstream.
Partially revert the commit mentioned in the Fixes line to make sure that
allocation and erasing multicast struct are locked.
BUG: KASAN: use-after-free in ucma_cleanup_multicast drivers/infiniband/core/ucma.c:491 [inline]
BUG: KASAN: use-after-free in ucma_destroy_private_ctx+0x914/0xb70 drivers/infiniband/core/ucma.c:579
Read of size 8 at addr ffff88801bb74b00 by task syz-executor.1/25529
CPU: 0 PID: 25529 Comm: syz-executor.1 Not tainted 5.16.0-rc7-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
print_address_description.constprop.0.cold+0x8d/0x320 mm/kasan/report.c:247
__kasan_report mm/kasan/report.c:433 [inline]
kasan_report.cold+0x83/0xdf mm/kasan/report.c:450
ucma_cleanup_multicast drivers/infiniband/core/ucma.c:491 [inline]
ucma_destroy_private_ctx+0x914/0xb70 drivers/infiniband/core/ucma.c:579
ucma_destroy_id+0x1e6/0x280 drivers/infiniband/core/ucma.c:614
ucma_write+0x25c/0x350 drivers/infiniband/core/ucma.c:1732
vfs_write+0x28e/0xae0 fs/read_write.c:588
ksys_write+0x1ee/0x250 fs/read_write.c:643
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x44/0xae
Currently the xarray search can touch a concurrently freeing mc as the
xa_for_each() is not surrounded by any lock. Rather than hold the lock for
a full scan hold it only for the effected items, which is usually an empty
list.
Fixes: 95fe51096b ("RDMA/ucma: Remove mc_list and rely on xarray")
Link: https://lore.kernel.org/r/1cda5fabb1081e8d16e39a48d3a4f8160cea88b8.1642491047.git.leonro@nvidia.com
Reported-by: syzbot+e3f96c43d19782dd14a7@syzkaller.appspotmail.com
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 20679094a0161c94faf77e373fa3f7428a8e14bd ]
Currently, when cma_resolve_ib_dev() searches for a matching GID it will
stop searching after encountering the first empty GID table entry. This
behavior is wrong since neither IB nor RoCE spec enforce tightly packed
GID tables.
For example, when the matching valid GID entry exists at index N, and if a
GID entry is empty at index N-1, cma_resolve_ib_dev() will fail to find
the matching valid entry.
Fix it by making cma_resolve_ib_dev() continue searching even after
encountering missing entries.
Fixes: f17df3b0de ("RDMA/cma: Add support for AF_IB to rdma_resolve_addr()")
Link: https://lore.kernel.org/r/b7346307e3bb396c43d67d924348c6c496493991.1639055490.git.leonro@nvidia.com
Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Reviewed-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 483d805191a23191f8294bbf9b4e94836f5d92e4 ]
Currently, ib_find_gid() will stop searching after encountering the first
empty GID table entry. This behavior is wrong since neither IB nor RoCE
spec enforce tightly packed GID tables.
For example, when a valid GID entry exists at index N, and if a GID entry
is empty at index N-1, ib_find_gid() will fail to find the valid entry.
Fix it by making ib_find_gid() continue searching even after encountering
missing entries.
Fixes: 5eb620c81c ("IB/core: Add helpers for uncached GID and P_Key searches")
Link: https://lore.kernel.org/r/e55d331b96cecfc2cf19803d16e7109ea966882d.1639055490.git.leonro@nvidia.com
Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Reviewed-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 7694a7de22c53a312ea98960fcafc6ec62046531 upstream.
Because of the possible failure of the allocation, data might be NULL
pointer and will cause the dereference of the NULL pointer later.
Therefore, it might be better to check it and return -ENOMEM.
Fixes: 6884c6c4bd ("RDMA/verbs: Store the write/write_ex uapi entry points in the uverbs_api")
Link: https://lore.kernel.org/r/20211231093315.1917667-1-jiasheng@iscas.ac.cn
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b35a0f4dd544eaa6162b6d2f13a2557a121ae5fd upstream.
If dst->is_global field is not set, the GRH fields are not cleared
and the following infoleak is reported.
=====================================================
BUG: KMSAN: kernel-infoleak in instrument_copy_to_user include/linux/instrumented.h:121 [inline]
BUG: KMSAN: kernel-infoleak in _copy_to_user+0x1c9/0x270 lib/usercopy.c:33
instrument_copy_to_user include/linux/instrumented.h:121 [inline]
_copy_to_user+0x1c9/0x270 lib/usercopy.c:33
copy_to_user include/linux/uaccess.h:209 [inline]
ucma_init_qp_attr+0x8c7/0xb10 drivers/infiniband/core/ucma.c:1242
ucma_write+0x637/0x6c0 drivers/infiniband/core/ucma.c:1732
vfs_write+0x8ce/0x2030 fs/read_write.c:588
ksys_write+0x28b/0x510 fs/read_write.c:643
__do_sys_write fs/read_write.c:655 [inline]
__se_sys_write fs/read_write.c:652 [inline]
__ia32_sys_write+0xdb/0x120 fs/read_write.c:652
do_syscall_32_irqs_on arch/x86/entry/common.c:114 [inline]
__do_fast_syscall_32+0x96/0xf0 arch/x86/entry/common.c:180
do_fast_syscall_32+0x34/0x70 arch/x86/entry/common.c:205
do_SYSENTER_32+0x1b/0x20 arch/x86/entry/common.c:248
entry_SYSENTER_compat_after_hwframe+0x4d/0x5c
Local variable resp created at:
ucma_init_qp_attr+0xa4/0xb10 drivers/infiniband/core/ucma.c:1214
ucma_write+0x637/0x6c0 drivers/infiniband/core/ucma.c:1732
Bytes 40-59 of 144 are uninitialized
Memory access of size 144 starts at ffff888167523b00
Data copied to user address 0000000020000100
CPU: 1 PID: 25910 Comm: syz-executor.1 Not tainted 5.16.0-rc5-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
=====================================================
Fixes: 4ba66093bd ("IB/core: Check for global flag when using ah_attr")
Link: https://lore.kernel.org/r/0e9dd51f93410b7b2f4f5562f52befc878b71afa.1641298868.git.leonro@nvidia.com
Reported-by: syzbot+6d532fa8f9463da290bc@syzkaller.appspotmail.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 6cd7397d01c4a3e09757840299e4f114f0aa5fa0 ]
Preset both receive and send CQ pointers prior to call to the drivers and
overwrite it later again till the mlx4 is going to be changed do not
overwrite ibqp properties.
This change is needed for mlx5, because in case of QP creation failure, it
will go to the path of QP destroy which relies on proper CQ pointers.
BUG: KASAN: use-after-free in create_qp.cold+0x164/0x16e [mlx5_ib]
Write of size 8 at addr ffff8880064c55c0 by task a.out/246
CPU: 0 PID: 246 Comm: a.out Not tainted 5.15.0+ #291
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
Call Trace:
dump_stack_lvl+0x45/0x59
print_address_description.constprop.0+0x1f/0x140
kasan_report.cold+0x83/0xdf
create_qp.cold+0x164/0x16e [mlx5_ib]
mlx5_ib_create_qp+0x358/0x28a0 [mlx5_ib]
create_qp.part.0+0x45b/0x6a0 [ib_core]
ib_create_qp_user+0x97/0x150 [ib_core]
ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x92c/0x1250 [ib_uverbs]
ib_uverbs_cmd_verbs+0x1c38/0x3150 [ib_uverbs]
ib_uverbs_ioctl+0x169/0x260 [ib_uverbs]
__x64_sys_ioctl+0x866/0x14d0
do_syscall_64+0x3d/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
Allocated by task 246:
kasan_save_stack+0x1b/0x40
__kasan_kmalloc+0xa4/0xd0
create_qp.part.0+0x92/0x6a0 [ib_core]
ib_create_qp_user+0x97/0x150 [ib_core]
ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x92c/0x1250 [ib_uverbs]
ib_uverbs_cmd_verbs+0x1c38/0x3150 [ib_uverbs]
ib_uverbs_ioctl+0x169/0x260 [ib_uverbs]
__x64_sys_ioctl+0x866/0x14d0
do_syscall_64+0x3d/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
Freed by task 246:
kasan_save_stack+0x1b/0x40
kasan_set_track+0x1c/0x30
kasan_set_free_info+0x20/0x30
__kasan_slab_free+0x10c/0x150
slab_free_freelist_hook+0xb4/0x1b0
kfree+0xe7/0x2a0
create_qp.part.0+0x52b/0x6a0 [ib_core]
ib_create_qp_user+0x97/0x150 [ib_core]
ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x92c/0x1250 [ib_uverbs]
ib_uverbs_cmd_verbs+0x1c38/0x3150 [ib_uverbs]
ib_uverbs_ioctl+0x169/0x260 [ib_uverbs]
__x64_sys_ioctl+0x866/0x14d0
do_syscall_64+0x3d/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
Fixes: 514aee660d ("RDMA: Globally allocate and release QP memory")
Link: https://lore.kernel.org/r/2dbb2e2cbb1efb188a500e5634be1d71956424ce.1636631035.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 911a81c9c7092bfd75432ce79b2ef879127ea065 ]
The 'struct attribute' flex array contains some struct lock_class_key's
which become big when lockdep is turned on. Big enough that some drivers
will not load when CONFIG_PROVE_LOCKING=y because they cannot allocate
enough memory:
WARNING: CPU: 36 PID: 8 at mm/page_alloc.c:5350 __alloc_pages+0x27e/0x3e0
Call Trace:
kmalloc_order+0x2a/0xb0
kmalloc_order_trace+0x19/0xf0
__kmalloc+0x231/0x270
ib_setup_port_attrs+0xd8/0x870 [ib_core]
ib_register_device+0x419/0x4e0 [ib_core]
bnxt_re_task+0x208/0x2d0 [bnxt_re]
Link: https://lore.kernel.org/r/20211019002656.17745-1-wangyugui@e16-tech.com
Signed-off-by: wangyugui <wangyugui@e16-tech.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f1a090f09f42be5a5542009f0be310fdb3e768fc ]
If the driver returns a new MR during rereg it has to fill it with the
IOVA from the proper source. If IB_MR_REREG_TRANS is set then the IOVA is
cmd.hca_va, otherwise the IOVA comes from the old MR. mlx5 for example has
two calls inside rereg_mr:
return create_real_mr(new_pd, umem, mr->ibmr.iova,
new_access_flags);
and
return create_real_mr(new_pd, new_umem, iova, new_access_flags);
Unconditionally overwriting the iova in the newly allocated MR will
corrupt the iova if the first path is used.
Remove the redundant initializations from ib_uverbs_rereg_mr().
Fixes: 6e0954b11c ("RDMA/uverbs: Allow drivers to create a new HW object during rereg_mr")
Link: https://lore.kernel.org/r/4b0a31bbc372842613286a10d7a8cbb0ee6069c7.1635400472.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
The FSM can run in a circle allowing rdma_resolve_ip() to be called twice
on the same id_priv. While this cannot happen without going through the
work, it violates the invariant that the same address resolution
background request cannot be active twice.
CPU 1 CPU 2
rdma_resolve_addr():
RDMA_CM_IDLE -> RDMA_CM_ADDR_QUERY
rdma_resolve_ip(addr_handler) #1
process_one_req(): for #1
addr_handler():
RDMA_CM_ADDR_QUERY -> RDMA_CM_ADDR_BOUND
mutex_unlock(&id_priv->handler_mutex);
[.. handler still running ..]
rdma_resolve_addr():
RDMA_CM_ADDR_BOUND -> RDMA_CM_ADDR_QUERY
rdma_resolve_ip(addr_handler)
!! two requests are now on the req_list
rdma_destroy_id():
destroy_id_handler_unlock():
_destroy_id():
cma_cancel_operation():
rdma_addr_cancel()
// process_one_req() self removes it
spin_lock_bh(&lock);
cancel_delayed_work(&req->work);
if (!list_empty(&req->list)) == true
! rdma_addr_cancel() returns after process_on_req #1 is done
kfree(id_priv)
process_one_req(): for #2
addr_handler():
mutex_lock(&id_priv->handler_mutex);
!! Use after free on id_priv
rdma_addr_cancel() expects there to be one req on the list and only
cancels the first one. The self-removal behavior of the work only happens
after the handler has returned. This yields a situations where the
req_list can have two reqs for the same "handle" but rdma_addr_cancel()
only cancels the first one.
The second req remains active beyond rdma_destroy_id() and will
use-after-free id_priv once it inevitably triggers.
Fix this by remembering if the id_priv has called rdma_resolve_ip() and
always cancel before calling it again. This ensures the req_list never
gets more than one item in it and doesn't cost anything in the normal flow
that never uses this strange error path.
Link: https://lore.kernel.org/r/0-v1-3bc675b8006d+22-syz_cancel_uaf_jgg@nvidia.com
Cc: stable@vger.kernel.org
Fixes: e51060f08a ("IB: IP address based RDMA connection manager")
Reported-by: syzbot+dc3dfba010d7671e05f5@syzkaller.appspotmail.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
If the state is not idle then rdma_bind_addr() will immediately fail and
no change to global state should happen.
For instance if the state is already RDMA_CM_LISTEN then this will corrupt
the src_addr and would cause the test in cma_cancel_operation():
if (cma_any_addr(cma_src_addr(id_priv)) && !id_priv->cma_dev)
To view a mangled src_addr, eg with a IPv6 loopback address but an IPv4
family, failing the test.
This would manifest as this trace from syzkaller:
BUG: KASAN: use-after-free in __list_add_valid+0x93/0xa0 lib/list_debug.c:26
Read of size 8 at addr ffff8881546491e0 by task syz-executor.1/32204
CPU: 1 PID: 32204 Comm: syz-executor.1 Not tainted 5.12.0-rc8-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:79 [inline]
dump_stack+0x141/0x1d7 lib/dump_stack.c:120
print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:232
__kasan_report mm/kasan/report.c:399 [inline]
kasan_report.cold+0x7c/0xd8 mm/kasan/report.c:416
__list_add_valid+0x93/0xa0 lib/list_debug.c:26
__list_add include/linux/list.h:67 [inline]
list_add_tail include/linux/list.h:100 [inline]
cma_listen_on_all drivers/infiniband/core/cma.c:2557 [inline]
rdma_listen+0x787/0xe00 drivers/infiniband/core/cma.c:3751
ucma_listen+0x16a/0x210 drivers/infiniband/core/ucma.c:1102
ucma_write+0x259/0x350 drivers/infiniband/core/ucma.c:1732
vfs_write+0x28e/0xa30 fs/read_write.c:603
ksys_write+0x1ee/0x250 fs/read_write.c:658
do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
entry_SYSCALL_64_after_hwframe+0x44/0xae
Which is indicating that an rdma_id_private was destroyed without doing
cma_cancel_listens().
Instead of trying to re-use the src_addr memory to indirectly create an
any address build one explicitly on the stack and bind to that as any
other normal flow would do.
Link: https://lore.kernel.org/r/0-v1-9fbb33f5e201+2a-cma_listen_jgg@nvidia.com
Cc: stable@vger.kernel.org
Fixes: 732d41c545 ("RDMA/cma: Make the locking for automatic state transition more clear")
Reported-by: syzbot+6bb0528b13611047209c@syzkaller.appspotmail.com
Tested-by: Hao Sun <sunhao.th@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
If cma_listen_on_all() fails it leaves the per-device ID still on the
listen_list but the state is not set to RDMA_CM_ADDR_BOUND.
When the cmid is eventually destroyed cma_cancel_listens() is not called
due to the wrong state, however the per-device IDs are still holding the
refcount preventing the ID from being destroyed, thus deadlocking:
task:rping state:D stack: 0 pid:19605 ppid: 47036 flags:0x00000084
Call Trace:
__schedule+0x29a/0x780
? free_unref_page_commit+0x9b/0x110
schedule+0x3c/0xa0
schedule_timeout+0x215/0x2b0
? __flush_work+0x19e/0x1e0
wait_for_completion+0x8d/0xf0
_destroy_id+0x144/0x210 [rdma_cm]
ucma_close_id+0x2b/0x40 [rdma_ucm]
__destroy_id+0x93/0x2c0 [rdma_ucm]
? __xa_erase+0x4a/0xa0
ucma_destroy_id+0x9a/0x120 [rdma_ucm]
ucma_write+0xb8/0x130 [rdma_ucm]
vfs_write+0xb4/0x250
ksys_write+0xb5/0xd0
? syscall_trace_enter.isra.19+0x123/0x190
do_syscall_64+0x33/0x40
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Ensure that cma_listen_on_all() atomically unwinds its action under the
lock during error.
Fixes: c80a0c52d8 ("RDMA/cma: Add missing error handling of listen_id")
Link: https://lore.kernel.org/r/20210913093344.17230-1-thomas.liu@ucloud.cn
Signed-off-by: Tao Liu <thomas.liu@ucloud.cn>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
ROCE uses IGMP for Multicast instead of the native Infiniband system where
joins are required in order to post messages on the Multicast group. On
Ethernet one can send Multicast messages to arbitrary addresses without
the need to subscribe to a group.
So ROCE correctly does not send IGMP joins during rdma_join_multicast().
F.e. in cma_iboe_join_multicast() we see:
if (addr->sa_family == AF_INET) {
if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) {
ib.rec.hop_limit = IPV6_DEFAULT_HOPLIMIT;
if (!send_only) {
err = cma_igmp_send(ndev, &ib.rec.mgid,
true);
}
}
} else {
So the IGMP join is suppressed as it is unnecessary.
However no such check is done in destroy_mc(). And therefore leaving a
sendonly multicast group will send an IGMP leave.
This means that the following scenario can lead to a multicast receiver
unexpectedly being unsubscribed from a MC group:
1. Sender thread does a sendonly join on MC group X. No IGMP join
is sent.
2. Receiver thread does a regular join on the same MC Group x.
IGMP join is sent and the receiver begins to get messages.
3. Sender thread terminates and destroys MC group X.
IGMP leave is sent and the receiver no longer receives data.
This patch adds the same logic for sendonly joins to destroy_mc() that is
also used in cma_iboe_join_multicast().
Fixes: ab15c95a17 ("IB/core: Support for CMA multicast join flags")
Link: https://lore.kernel.org/r/alpine.DEB.2.22.394.2109081340540.668072@gentwo.de
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Pull rdma updates from Jason Gunthorpe:
"This is quite a small cycle, no major series stands out. The HNS and
rxe drivers saw the most activity this cycle, with rxe being broken
for a good chunk of time. The significant deleted line count is due to
a SPDX cleanup series.
Summary:
- Various cleanup and small features for rtrs
- kmap_local_page() conversions
- Driver updates and fixes for: efa, rxe, mlx5, hfi1, qed, hns
- Cache the IB subnet prefix
- Rework how CRC is calcuated in rxe
- Clean reference counting in iwpm's netlink
- Pull object allocation and lifecycle for user QPs to the uverbs
core code
- Several small hns features and continued general code cleanups
- Fix the scatterlist confusion of orig_nents/nents introduced in an
earlier patch creating the append operation"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (90 commits)
RDMA/mlx5: Relax DCS QP creation checks
RDMA/hns: Delete unnecessary blank lines.
RDMA/hns: Encapsulate the qp db as a function
RDMA/hns: Adjust the order in which irq are requested and enabled
RDMA/hns: Remove RST2RST error prints for hw v1
RDMA/hns: Remove dqpn filling when modify qp from Init to Init
RDMA/hns: Fix QP's resp incomplete assignment
RDMA/hns: Fix query destination qpn
RDMA/hfi1: Convert to SPDX identifier
IB/rdmavt: Convert to SPDX identifier
RDMA/hns: Bugfix for incorrect association between dip_idx and dgid
RDMA/hns: Bugfix for the missing assignment for dip_idx
RDMA/hns: Bugfix for data type of dip_idx
RDMA/hns: Fix incorrect lsn field
RDMA/irdma: Remove the repeated declaration
RDMA/core/sa_query: Retry SA queries
RDMA: Use the sg_table directly and remove the opencoded version from umem
lib/scatterlist: Fix wrong update of orig_nents
lib/scatterlist: Provide a dedicated function to support table append
RDMA/hns: Delete unused hns bitmap interface
...
From Maor Gottlieb
====================
Fix the use of nents and orig_nents in the sg table append helpers. The
nents should be used by the DMA layer to store the number of DMA mapped
sges, the orig_nents is the number of CPU sges.
Since the sg append logic doesn't always create a SGL with exactly
orig_nents entries store a total_nents as well to allow the table to be
properly free'd and reorganize the freeing logic to share across all the
use cases.
====================
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
* 'sg_nents':
RDMA: Use the sg_table directly and remove the opencoded version from umem
lib/scatterlist: Fix wrong update of orig_nents
lib/scatterlist: Provide a dedicated function to support table append
A MAD packet is sent as an unreliable datagram (UD). SA requests are sent
as MAD packets. As such, SA requests or responses may be silently dropped.
IB Core's MAD layer has a timeout and retry mechanism, which amongst
other, is used by RDMA CM. But it is not used by SA queries. The lack of
retries of SA queries leads to long specified timeout, and error being
returned in case of packet loss. The ULP or user-land process has to
perform the retry.
Fix this by taking advantage of the MAD layer's retry mechanism.
First, a check against a zero timeout is added in rdma_resolve_route(). In
send_mad(), we set the MAD layer timeout to one tenth of the specified
timeout and the number of retries to 10. The special case when timeout is
less than 10 is handled.
With this fix:
# ucmatose -c 1000 -S 1024 -C 1
runs stable on an Infiniband fabric. Without this fix, we see an
intermittent behavior and it errors out with:
cmatose: event: RDMA_CM_EVENT_ROUTE_ERROR, error: -110
(110 is ETIMEDOUT)
Link: https://lore.kernel.org/r/1628784755-28316-1-git-send-email-haakon.bugge@oracle.com
Signed-off-by: Håkon Bugge <haakon.bugge@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
orig_nents should represent the number of entries with pages,
but __sg_alloc_table_from_pages sets orig_nents as the number of
total entries in the table. This is wrong when the API is used for
dynamic allocation where not all the table entries are mapped with
pages. It wasn't observed until now, since RDMA umem who uses this
API in the dynamic form doesn't use orig_nents implicit or explicit
by the scatterlist APIs.
Fix it by changing the append API to track the SG append table
state and have an API to free the append table according to the
total number of entries in the table.
Now all APIs set orig_nents as number of enries with pages.
Fixes: 07da1223ec ("lib/scatterlist: Add support in dynamic allocation of SG table from pages")
Link: https://lore.kernel.org/r/20210824142531.3877007-3-maorg@nvidia.com
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
RDMA is the only in-kernel user that uses __sg_alloc_table_from_pages to
append pages dynamically. In the next patch. That mode will be extended
and that function will get more parameters. So separate it into a unique
function to make such change more clear.
Link: https://lore.kernel.org/r/20210824142531.3877007-2-maorg@nvidia.com
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
