linux/drivers/infiniband/core/core_priv.h

377 lines
11 KiB

IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA/core: Fix stale RoCE GIDs during netdev events at registration RoCE GID entries become stale when netdev properties change during the IB device registration window. This is reproducible with a udev rule that sets a MAC address when a VF netdev appears: ACTION=="add", SUBSYSTEM=="net", KERNEL=="eth4", \ RUN+="/sbin/ip link set eth4 address 88:22:33:44:55:66" After VF creation, show_gids displays GIDs derived from the original random MAC rather than the configured one. The root cause is a race between netdev event processing and device registration: CPU 0 (driver) CPU 1 (udev/workqueue) ────────────── ────────────────────── ib_register_device() ib_cache_setup_one() gid_table_setup_one() _gid_table_setup_one() ← GID table allocated rdma_roce_rescan_device() ← GIDs populated with OLD MAC ip link set eth4 addr NEW_MAC NETDEV_CHANGEADDR queued netdevice_event_work_handler() ib_enum_all_roce_netdevs() ← Iterates DEVICE_REGISTERED ← Device NOT marked yet, SKIP! enable_device_and_get() xa_set_mark(DEVICE_REGISTERED) ← Too late, event was lost The netdev event handler uses ib_enum_all_roce_netdevs() which only iterates devices marked DEVICE_REGISTERED. However, this mark is set late in the registration process, after the GID cache is already populated. Events arriving in this window are silently dropped. Fix this by introducing a new xarray mark DEVICE_GID_UPDATES that is set immediately after the GID table is allocated and initialized. Use the new mark in ib_enum_all_roce_netdevs() function to iterate devices instead of DEVICE_REGISTERED. This is safe because: - After _gid_table_setup_one(), all required structures exist (port_data, immutable, cache.gid) - The GID table mutex serializes concurrent access between the initial rescan and event handlers - Event handlers correctly update stale GIDs even when racing with rescan - The mark is cleared in ib_cache_cleanup_one() before teardown This also fixes similar races for IP address events (inetaddr_event, inet6addr_event) which use the same enumeration path. Fixes: 0df91bb67334 ("RDMA/devices: Use xarray to store the client_data") Signed-off-by: Jiri Pirko <jiri@nvidia.com> Link: https://patch.msgid.link/20260127093839.126291-1-jiri@resnulli.us Reported-by: syzbot+881d65229ca4f9ae8c84@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=881d65229ca4f9ae8c84 Signed-off-by: Leon Romanovsky <leon@kernel.org>
2026-01-27 10:38:39 +01:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
IB/core: Add RoCE GID table management RoCE GIDs are based on IP addresses configured on Ethernet net-devices which relate to the RDMA (RoCE) device port. Currently, each of the low-level drivers that support RoCE (ocrdma, mlx4) manages its own RoCE port GID table. As there's nothing which is essentially vendor specific, we generalize that, and enhance the RDMA core GID cache to do this job. In order to populate the GID table, we listen for events: (a) netdev up/down/change_addr events - if a netdev is built onto our RoCE device, we need to add/delete its IPs. This involves adding all GIDs related to this ndev, add default GIDs, etc. (b) inet events - add new GIDs (according to the IP addresses) to the table. For programming the port RoCE GID table, providers must implement the add_gid and del_gid callbacks. RoCE GID management requires us to state the associated net_device alongside the GID. This information is necessary in order to manage the GID table. For example, when a net_device is removed, its associated GIDs need to be removed as well. RoCE mandates generating a default GID for each port, based on the related net-device's IPv6 link local. In contrast to the GID based on the regular IPv6 link-local (as we generate GID per IP address), the default GID is also available when the net device is down (in order to support loopback). Locking is done as follows: The patch modify the GID table code both for new RoCE drivers implementing the add_gid/del_gid callbacks and for current RoCE and IB drivers that do not. The flows for updating the table are different, so the locking requirements are too. While updating RoCE GID table, protection against multiple writers is achieved via mutex_lock(&table->lock). Since writing to a table requires us to find an entry (possible a free entry) in the table and then modify it, this mutex protects both the find_gid and write_gid ensuring the atomicity of the action. Each entry in the GID cache is protected by rwlock. In RoCE, writing (usually results from netdev notifier) involves invoking the vendor's add_gid and del_gid callbacks, which could sleep. Therefore, an invalid flag is added for each entry. Updates for RoCE are done via a workqueue, thus sleeping is permitted. In IB, updates are done in write_lock_irq(&device->cache.lock), thus write_gid isn't allowed to sleep and add_gid/del_gid are not called. When passing net-device into/out-of the GID cache, the device is always passed held (dev_hold). The code uses a single work item for updating all RDMA devices, following a netdev or inet notifier. The patch moves the cache from being a client (which was incorrect, as the cache is part of the IB infrastructure) to being explicitly initialized/freed when a device is registered/removed. Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2015-07-30 18:33:26 +03:00
IB/core: Avoid deadlock during netlink message handling When rdmacm module is not loaded, and when netlink message is received to get char device info, it results into a deadlock due to recursive locking of rdma_nl_mutex with the below call sequence. [..] rdma_nl_rcv() mutex_lock() [..] rdma_nl_rcv_msg() ib_get_client_nl_info() request_module() iw_cm_init() rdma_nl_register() mutex_lock(); <- Deadlock, acquiring mutex again Due to above call sequence, following call trace and deadlock is observed. kernel: __mutex_lock+0x35e/0x860 kernel: ? __mutex_lock+0x129/0x860 kernel: ? rdma_nl_register+0x1a/0x90 [ib_core] kernel: rdma_nl_register+0x1a/0x90 [ib_core] kernel: ? 0xffffffffc029b000 kernel: iw_cm_init+0x34/0x1000 [iw_cm] kernel: do_one_initcall+0x67/0x2d4 kernel: ? kmem_cache_alloc_trace+0x1ec/0x2a0 kernel: do_init_module+0x5a/0x223 kernel: load_module+0x1998/0x1e10 kernel: ? __symbol_put+0x60/0x60 kernel: __do_sys_finit_module+0x94/0xe0 kernel: do_syscall_64+0x5a/0x270 kernel: entry_SYSCALL_64_after_hwframe+0x49/0xbe process stack trace: [<0>] __request_module+0x1c9/0x460 [<0>] ib_get_client_nl_info+0x5e/0xb0 [ib_core] [<0>] nldev_get_chardev+0x1ac/0x320 [ib_core] [<0>] rdma_nl_rcv_msg+0xeb/0x1d0 [ib_core] [<0>] rdma_nl_rcv+0xcd/0x120 [ib_core] [<0>] netlink_unicast+0x179/0x220 [<0>] netlink_sendmsg+0x2f6/0x3f0 [<0>] sock_sendmsg+0x30/0x40 [<0>] ___sys_sendmsg+0x27a/0x290 [<0>] __sys_sendmsg+0x58/0xa0 [<0>] do_syscall_64+0x5a/0x270 [<0>] entry_SYSCALL_64_after_hwframe+0x49/0xbe To overcome this deadlock and to allow multiple netlink messages to progress in parallel, following scheme is implemented. 1. Split the lock protecting the cb_table into a per-index lock, and make it a rwlock. This lock is used to ensure no callbacks are running after unregistration returns. Since a module will not be registered once it is already running callbacks, this avoids the deadlock. 2. Use smp_store_release() to update the cb_table during registration so that no lock is required. This avoids lockdep problems with thinking all the rwsems are the same lock class. Fixes: 0e2d00eb6fd45 ("RDMA: Add NLDEV_GET_CHARDEV to allow char dev discovery and autoload") Link: https://lore.kernel.org/r/20191015080733.18625-1-leon@kernel.org Signed-off-by: Parav Pandit <parav@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2019-10-15 11:07:33 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
RDMA: Support more than 255 rdma ports Current code uses many different types when dealing with a port of a RDMA device: u8, unsigned int and u32. Switch to u32 to clean up the logic. This allows us to make (at least) the core view consistent and use the same type. Unfortunately not all places can be converted. Many uverbs functions expect port to be u8 so keep those places in order not to break UAPIs. HW/Spec defined values must also not be changed. With the switch to u32 we now can support devices with more than 255 ports. U32_MAX is reserved to make control logic a bit easier to deal with. As a device with U32_MAX ports probably isn't going to happen any time soon this seems like a non issue. When a device with more than 255 ports is created uverbs will report the RDMA device as having 255 ports as this is the max currently supported. The verbs interface is not changed yet because the IBTA spec limits the port size in too many places to be u8 and all applications that relies in verbs won't be able to cope with this change. At this stage, we are extending the interfaces that are using vendor channel solely Once the limitation is lifted mlx5 in switchdev mode will be able to have thousands of SFs created by the device. As the only instance of an RDMA device that reports more than 255 ports will be a representor device and it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other ULPs aren't effected by this change and their sysfs/interfaces that are exposes to userspace can remain unchanged. While here cleanup some alignment issues and remove unneeded sanity checks (mainly in rdmavt), Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.org Signed-off-by: Mark Bloch <mbloch@nvidia.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-03-01 09:04:20 +02:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
IB/core: Enforce PKey security on QPs Add new LSM hooks to allocate and free security contexts and check for permission to access a PKey. Allocate and free a security context when creating and destroying a QP. This context is used for controlling access to PKeys. When a request is made to modify a QP that changes the port, PKey index, or alternate path, check that the QP has permission for the PKey in the PKey table index on the subnet prefix of the port. If the QP is shared make sure all handles to the QP also have access. Store which port and PKey index a QP is using. After the reset to init transition the user can modify the port, PKey index and alternate path independently. So port and PKey settings changes can be a merge of the previous settings and the new ones. In order to maintain access control if there are PKey table or subnet prefix change keep a list of all QPs are using each PKey index on each port. If a change occurs all QPs using that device and port must have access enforced for the new cache settings. These changes add a transaction to the QP modify process. Association with the old port and PKey index must be maintained if the modify fails, and must be removed if it succeeds. Association with the new port and PKey index must be established prior to the modify and removed if the modify fails. 1. When a QP is modified to a particular Port, PKey index or alternate path insert that QP into the appropriate lists. 2. Check permission to access the new settings. 3. If step 2 grants access attempt to modify the QP. 4a. If steps 2 and 3 succeed remove any prior associations. 4b. If ether fails remove the new setting associations. If a PKey table or subnet prefix changes walk the list of QPs and check that they have permission. If not send the QP to the error state and raise a fatal error event. If it's a shared QP make sure all the QPs that share the real_qp have permission as well. If the QP that owns a security structure is denied access the security structure is marked as such and the QP is added to an error_list. Once the moving the QP to error is complete the security structure mark is cleared. Maintaining the lists correctly turns QP destroy into a transaction. The hardware driver for the device frees the ib_qp structure, so while the destroy is in progress the ib_qp pointer in the ib_qp_security struct is undefined. When the destroy process begins the ib_qp_security structure is marked as destroying. This prevents any action from being taken on the QP pointer. After the QP is destroyed successfully it could still listed on an error_list wait for it to be processed by that flow before cleaning up the structure. If the destroy fails the QPs port and PKey settings are reinserted into the appropriate lists, the destroying flag is cleared, and access control is enforced, in case there were any cache changes during the destroy flow. To keep the security changes isolated a new file is used to hold security related functionality. Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> [PM: merge fixup in ib_verbs.h and uverbs_cmd.c] Signed-off-by: Paul Moore <paul@paul-moore.com>
2017-05-19 15:48:52 +03:00
RDMA/cm: Move debug counters to be under relevant IB device The sysfs layout is created by CM incorrectly presented RDMA devices with InfiniBand link layer. Layout of such devices represents device tree of connections. By moving CM statistics to be under relevant port of IB device, we will fix the following issues: * Symlink name - It used device name instead of specific identifier. * Target location - It was supposed to point to PCI-ID/infiniband_cm/ instead of PCI-ID/infiniband/ * Target name - It created extra device file under already existing device folder, e.g. mlx5_0/mlx5_0 * Crash during boot with RDMA persistent naming patches. sysfs: cannot create duplicate filename '/class/infiniband_cm/mlx5_0' CPU: 29 PID: 433 Comm: modprobe Not tainted 5.0.0-rc5+ #178 Call Trace: dump_stack+0xcc/0x180 sysfs_warn_dup.cold.3+0x17/0x2d sysfs_do_create_link_sd.isra.2+0xd0/0xf0 device_add+0x7cb/0x1450 device_create_groups_vargs+0x1ae/0x220 device_create+0x93/0xc0 cm_add_one+0x38f/0xf60 [ib_cm] add_client_context+0x167/0x210 [ib_core] enable_device_and_get+0x230/0x3f0 [ib_core] ib_register_device+0x823/0xbf0 [ib_core] __mlx5_ib_add+0x45/0x150 [mlx5_ib] mlx5_ib_add+0x1b3/0x5e0 [mlx5_ib] mlx5_add_device+0x130/0x3a0 [mlx5_core] mlx5_register_interface+0x1a9/0x270 [mlx5_core] do_one_initcall+0x14f/0x5de do_init_module+0x247/0x7c0 load_module+0x4c2f/0x60d0 entry_SYSCALL_64_after_hwframe+0x49/0xbe After this change: [leonro@server ~]$ ls -al /sys/class/infiniband/ibp0s12f0/ports/1/ drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_rx_duplicates drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_rx_msgs drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_tx_msgs drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_tx_retries Fixes: 110cf374a809 ("infiniband: make cm_device use a struct device and not a kobject.") Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2019-03-11 14:40:31 +02:00
RDMA/cm: Move debug counters to be under relevant IB device The sysfs layout is created by CM incorrectly presented RDMA devices with InfiniBand link layer. Layout of such devices represents device tree of connections. By moving CM statistics to be under relevant port of IB device, we will fix the following issues: * Symlink name - It used device name instead of specific identifier. * Target location - It was supposed to point to PCI-ID/infiniband_cm/ instead of PCI-ID/infiniband/ * Target name - It created extra device file under already existing device folder, e.g. mlx5_0/mlx5_0 * Crash during boot with RDMA persistent naming patches. sysfs: cannot create duplicate filename '/class/infiniband_cm/mlx5_0' CPU: 29 PID: 433 Comm: modprobe Not tainted 5.0.0-rc5+ #178 Call Trace: dump_stack+0xcc/0x180 sysfs_warn_dup.cold.3+0x17/0x2d sysfs_do_create_link_sd.isra.2+0xd0/0xf0 device_add+0x7cb/0x1450 device_create_groups_vargs+0x1ae/0x220 device_create+0x93/0xc0 cm_add_one+0x38f/0xf60 [ib_cm] add_client_context+0x167/0x210 [ib_core] enable_device_and_get+0x230/0x3f0 [ib_core] ib_register_device+0x823/0xbf0 [ib_core] __mlx5_ib_add+0x45/0x150 [mlx5_ib] mlx5_ib_add+0x1b3/0x5e0 [mlx5_ib] mlx5_add_device+0x130/0x3a0 [mlx5_core] mlx5_register_interface+0x1a9/0x270 [mlx5_core] do_one_initcall+0x14f/0x5de do_init_module+0x247/0x7c0 load_module+0x4c2f/0x60d0 entry_SYSCALL_64_after_hwframe+0x49/0xbe After this change: [leonro@server ~]$ ls -al /sys/class/infiniband/ibp0s12f0/ports/1/ drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_rx_duplicates drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_rx_msgs drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_tx_msgs drwxr-xr-x 2 root root 0 Mar 11 11:17 cm_tx_retries Fixes: 110cf374a809 ("infiniband: make cm_device use a struct device and not a kobject.") Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2019-03-11 14:40:31 +02:00
RDMA/core: Clean up cq pool mechanism The CQ pool mechanism had two problems: 1. The CQ pool lists were uninitialized in the device registration error flow. As a result, all the list pointers remained NULL. This caused the kernel to crash (in procedure ib_cq_pool_destroy) when that error flow was taken (and unregister called). The stack trace snippet: BUG: kernel NULL pointer dereference, address: 0000000000000000 #PF: supervisor read access in kernel mode #PF: error_code(0×0000) ? not-present page PGD 0 P4D 0 Oops: 0000 [#1] SMP PTI . . . RIP: 0010:ib_cq_pool_destroy+0x1b/0×70 [ib_core] . . . Call Trace: disable_device+0x9f/0×130 [ib_core] __ib_unregister_device+0x35/0×90 [ib_core] ib_register_device+0x529/0×610 [ib_core] __mlx5_ib_add+0x3a/0×70 [mlx5_ib] mlx5_add_device+0x87/0×1c0 [mlx5_core] mlx5_register_interface+0x74/0xc0 [mlx5_core] do_one_initcall+0x4b/0×1f4 do_init_module+0x5a/0×223 load_module+0x1938/0×1d40 2. At device unregister, when cleaning up the cq pool, the cq's in the pool lists were freed, but the cq entries were left in the list. The fix for the first issue is to initialize the cq pool lists when the ib_device structure is allocated for a new device (in procedure _ib_alloc_device). The fix for the second problem is to delete cq entries from the pool lists when cleaning up the cq pool. In addition, procedure ib_cq_pool_destroy() is renamed to the more appropriate name ib_cq_pool_cleanup(). Fixes: 4aa1615268a8 ("RDMA/core: Fix ordering of CQ pool destruction") Link: https://lore.kernel.org/r/20201208073545.9723-2-leon@kernel.org Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2020-12-08 09:35:43 +02:00
RDMA/core: Introduce shared CQ pool API Allow a ULP to ask the core to provide a completion queue based on a least-used search on a per-device CQ pools. The device CQ pools grow in a lazy fashion when more CQs are requested. This feature reduces the amount of interrupts when using many QPs. Using shared CQs allows for more effcient completion handling. It also reduces the amount of overhead needed for CQ contexts. Test setup: Intel(R) Xeon(R) Platinum 8176M CPU @ 2.10GHz servers. Running NVMeoF 4KB read IOs over ConnectX-5EX across Spectrum switch. TX-depth = 32. The patch was applied in the nvme driver on both the target and initiator. Four controllers are accessed from each core. In the current test case we have exposed sixteen NVMe namespaces using four different subsystems (four namespaces per subsystem) from one NVM port. Each controller allocated X queues (RDMA QPs) and attached to Y CQs. Before this series we had X == Y, i.e for four controllers we've created total of 4X QPs and 4X CQs. In the shared case, we've created 4X QPs and only X CQs which means that we have four controllers that share a completion queue per core. Until fourteen cores there is no significant change in performance and the number of interrupts per second is less than a million in the current case. ================================================== |Cores|Current KIOPs |Shared KIOPs |improvement| |-----|---------------|--------------|-----------| |14 |2332 |2723 |16.7% | |-----|---------------|--------------|-----------| |20 |2086 |2712 |30% | |-----|---------------|--------------|-----------| |28 |1971 |2669 |35.4% | |================================================= |Cores|Current avg lat|Shared avg lat|improvement| |-----|---------------|--------------|-----------| |14 |767us |657us |14.3% | |-----|---------------|--------------|-----------| |20 |1225us |943us |23% | |-----|---------------|--------------|-----------| |28 |1816us |1341us |26.1% | ======================================================== |Cores|Current interrupts|Shared interrupts|improvement| |-----|------------------|-----------------|-----------| |14 |1.6M/sec |0.4M/sec |72% | |-----|------------------|-----------------|-----------| |20 |2.8M/sec |0.6M/sec |72.4% | |-----|------------------|-----------------|-----------| |28 |2.9M/sec |0.8M/sec |63.4% | ==================================================================== |Cores|Current 99.99th PCTL lat|Shared 99.99th PCTL lat|improvement| |-----|------------------------|-----------------------|-----------| |14 |67ms |6ms |90.9% | |-----|------------------------|-----------------------|-----------| |20 |5ms |6ms |-10% | |-----|------------------------|-----------------------|-----------| |28 |8.7ms |6ms |25.9% | |=================================================================== Performance improvement with sixteen disks (sixteen CQs per core) is comparable. Link: https://lore.kernel.org/r/1590568495-101621-3-git-send-email-yaminf@mellanox.com Signed-off-by: Yamin Friedman <yaminf@mellanox.com> Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com> Reviewed-by: Max Gurtovoy <maxg@mellanox.com> Reviewed-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2020-05-27 11:34:53 +03:00