- 09 Jun, 2023 28 commits
-
-
David Howells authored
Make TLS's sendmsg() support MSG_SPLICE_PAGES. This causes pages to be spliced from the source iterator if possible. This allows ->sendpage() to be replaced by something that can handle multiple multipage folios in a single transaction. Signed-off-by:
David Howells <dhowells@redhat.com> cc: Chuck Lever <chuck.lever@oracle.com> cc: Boris Pismenny <borisp@nvidia.com> cc: John Fastabend <john.fastabend@gmail.com> cc: Jens Axboe <axboe@kernel.dk> cc: Matthew Wilcox <willy@infradead.org> Signed-off-by:
Jakub Kicinski <kuba@kernel.org>
-
David Howells authored
splice_direct_to_actor() doesn't manage SPLICE_F_MORE correctly[1] - and, as a result, it incorrectly signals/fails to signal MSG_MORE when splicing to a socket. The problem I'm seeing happens when a short splice occurs because we got a short read due to hitting the EOF on a file: as the length read (read_len) is less than the remaining size to be spliced (len), SPLICE_F_MORE (and thus MSG_MORE) is set. The issue is that, for the moment, we have no way to know *why* the short read occurred and so can't make a good decision on whether we *should* keep MSG_MORE set. MSG_SENDPAGE_NOTLAST was added to work around this, but that is also set incorrectly under some circumstances - for example if a short read fills a single pipe_buffer, but the next read would return more (seqfile can do this). This was observed with the multi_chunk_sendfile tests in the tls kselftest program. Some of those tests would hang and time out when the last chunk of file was less than the sendfile request size: build/kselftest/net/tls -r tls.12_aes_gcm.multi_chunk_sendfile This has been observed before[2] and worked around in AF_TLS[3]. Fix this by making splice_direct_to_actor() always signal SPLICE_F_MORE if we haven't yet hit the requested operation size. SPLICE_F_MORE remains signalled if the user passed it in to splice() but otherwise gets cleared when we've read sufficient data to fulfill the request. If, however, we get a premature EOF from ->splice_read(), have sent at least one byte and SPLICE_F_MORE was not set by the caller, ->splice_eof() will be invoked. Signed-off-by:
-
David Howells authored
Allow splice to undo the effects of MSG_MORE after prematurely ending a splice/sendfile due to getting an EOF condition (->splice_read() returned 0) after splice had called sendmsg() with MSG_MORE set when the user didn't set MSG_MORE. Suggested-by:
Linus Torvalds <torvalds@linux-foundation.org> Link: https://lore.kernel.org/r/CAHk-=wh=V579PDYvkpnTobCLGczbgxpMgGmmhqiTyE34Cpi5Gg@mail.gmail.com/Signed-off-by:
-
David Howells authored
Allow splice to end a Chelsio TLS record after prematurely ending a splice/sendfile due to getting an EOF condition (->splice_read() returned 0) after splice had called sendmsg() with MSG_MORE set when the user didn't set MSG_MORE. Suggested-by:
-
David Howells authored
Allow splice to undo the effects of MSG_MORE after prematurely ending a splice/sendfile due to getting an EOF condition (->splice_read() returned 0) after splice had called sendmsg() with MSG_MORE set when the user didn't set MSG_MORE. For UDP, a pending packet will not be emitted if the socket is closed before it is flushed; with this change, it be flushed by ->splice_eof(). For TCP, it's not clear that MSG_MORE is actually effective. Suggested-by:
-
David Howells authored
Allow splice to end a TLS record after prematurely ending a splice/sendfile due to getting an EOF condition (->splice_read() returned 0) after splice had called TLS with a sendmsg() with MSG_MORE set when the user didn't set MSG_MORE. Suggested-by:
-
David Howells authored
Allow splice to end a TLS record after prematurely ending a splice/sendfile due to getting an EOF condition (->splice_read() returned 0) after splice had called TLS with a sendmsg() with MSG_MORE set when the user didn't set MSG_MORE. Suggested-by:
-
David Howells authored
Add an optional method, ->splice_eof(), to allow splice to indicate the premature termination of a splice to struct file_operations and struct proto_ops. This is called if sendfile() or splice() encounters all of the following conditions inside splice_direct_to_actor(): (1) the user did not set SPLICE_F_MORE (splice only), and (2) an EOF condition occurred (->splice_read() returned 0), and (3) we haven't read enough to fulfill the request (ie. len > 0 still), and (4) we have already spliced at least one byte. A further patch will modify the behaviour of SPLICE_F_MORE to always be passed to the actor if either the user set it or we haven't yet read sufficient data to fulfill the request. Suggested-by:
-
David Howells authored
Replace generic_splice_sendpage() + splice_from_pipe + pipe_to_sendpage() with a net-specific handler, splice_to_socket(), that calls sendmsg() with MSG_SPLICE_PAGES set instead of calling ->sendpage(). MSG_MORE is used to indicate if the sendmsg() is expected to be followed with more data. This allows multiple pipe-buffer pages to be passed in a single call in a BVEC iterator, allowing the processing to be pushed down to a loop in the protocol driver. This helps pave the way for passing multipage folios down too. Protocols that haven't been converted to handle MSG_SPLICE_PAGES yet should just ignore it and do a normal sendmsg() for now - although that may be a bit slower as it may copy everything. Signed-off-by:
-
David Howells authored
Allow MSG_SPLICE_PAGES to be specified to sendmsg() but treat it as normal sendmsg for now. This means the data will just be copied until MSG_SPLICE_PAGES is handled. Signed-off-by:
-
David Howells authored
It is necessary to allow MSG_SENDPAGE_* to be passed into ->sendmsg() to allow sendmsg(MSG_SPLICE_PAGES) to replace ->sendpage(). Unblocking them in the network protocol, however, allows these flags to be passed in by userspace too[1]. Fix this by marking MSG_SENDPAGE_NOPOLICY, MSG_SENDPAGE_NOTLAST and MSG_SENDPAGE_DECRYPTED as internal flags, which causes sendmsg() to object if they are passed to sendmsg() by userspace. Network protocol ->sendmsg() implementations can then allow them through. Note that it should be possible to remove MSG_SENDPAGE_NOTLAST once sendpage is removed as a whole slew of pages will be passed in in one go by splice through sendmsg, with MSG_MORE being set if it has more data waiting in the pipe. Signed-off-by:
-
Eric Dumazet authored
tcp_mtu_probe() is still copying payload from skbs in the write queue, using skb_copy_bits(), ignoring potential errors. Modern TCP stack wants to only deal with payload found in page frags, as this is a prereq for TCPDirect (host stack might not have access to the payload) Signed-off-by:
Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/r/20230607214113.1992947-1-edumazet@google.comSigned-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxJakub Kicinski authored
Saeed Mahameed says: ==================== mlx5-updates-2023-06-06 1) Support 4 ports VF LAG, part 2/2 2) Few extra trivial cleanup patches Shay Drory Says: ================ Support 4 ports VF LAG, part 2/2 This series continues the series[1] "Support 4 ports VF LAG, part1/2". This series adds support for 4 ports VF LAG (single FDB E-Switch). This series of patches refactoring LAG code that make assumptions about VF LAG supporting only two ports and then enable 4 ports VF LAG. Patch 1: - Fix for ib rep code Patches 2-5: - Refactors LAG layer. Patches 6-7: - Block LAG types which doesn't support 4 ports. Patch 8: - Enable 4 ports VF LAG. This series specifically allows HCAs with 4 ports to create a VF LAG with only 4 ports. It is not possible to create a VF LAG with 2 or 3 ports using HCAs that have 4 ports. Currently, the Merged E-Switch feature only supports HCAs with 2 ports. However, upcoming patches will introduce support for HCAs with 4 ports. In order to activate VF LAG a user can execute: devlink dev eswitch set pci/0000:08:00.0 mode switchdev devlink dev eswitch set pci/0000:08:00.1 mode switchdev devlink dev eswitch set pci/0000:08:00.2 mode switchdev devlink dev eswitch set pci/0000:08:00.3 mode switchdev ip link add name bond0 type bond ip link set dev bond0 type bond mode 802.3ad ip link set dev eth2 master bond0 ip link set dev eth3 master bond0 ip link set dev eth4 master bond0 ip link set dev eth5 master bond0 Where eth2, eth3, eth4 and eth5 are net-interfaces of pci/0000:08:00.0 pci/0000:08:00.1 pci/0000:08:00.2 pci/0000:08:00.3 respectively. User can verify LAG state and type via debugfs: /sys/kernel/debug/mlx5/0000\:08\:00.0/lag/state /sys/kernel/debug/mlx5/0000\:08\:00.0/lag/type [1] https://lore.kernel.org/netdev/20230601060118.154015-1-saeed@kernel.org/T/#mf1d2083780970ba277bfe721554d4925f03f36d1 ================ * tag 'mlx5-updates-2023-06-06' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux: net/mlx5e: simplify condition after napi budget handling change mlx5/core: E-Switch, Allocate ECPF vport if it's an eswitch manager net/mlx5: Skip inline mode check after mlx5_eswitch_enable_locked() failure net/mlx5e: TC, refactor access to hash key net/mlx5e: Remove RX page cache leftovers net/mlx5e: Expose catastrophic steering error counters net/mlx5: Enable 4 ports VF LAG net/mlx5: LAG, block multiport eswitch LAG in case ldev have more than 2 ports net/mlx5: LAG, block multipath LAG in case ldev have more than 2 ports net/mlx5: LAG, change mlx5_shared_fdb_supported() to static net/mlx5: LAG, generalize handling of shared FDB net/mlx5: LAG, check if all eswitches are paired for shared FDB {net/RDMA}/mlx5: introduce lag_for_each_peer RDMA/mlx5: Free second uplink ib port ==================== Link: https://lore.kernel.org/r/20230607210410.88209-1-saeed@kernel.orgSigned-off-by:
-
Justin Chen authored
The netlink version of set_wol checks for not supported wolopts and avoids setting wol when the correct wolopt is already set. If we do the same with the ioctl version then we can remove these checks from the driver layer. Reviewed-by:
Simon Horman <simon.horman@corigine.com> Signed-off-by:
Justin Chen <justin.chen@broadcom.com> Reviewed-by:
Florian Fainelli <florian.fainelli@broadcom.com> Link: https://lore.kernel.org/r/1686179653-29750-1-git-send-email-justin.chen@broadcom.comSigned-off-by:
-
Jakub Kicinski authored
Russell King says: ==================== complete Lynx mdio device handling This series completes the mdio device lifetime handling for Lynx PCS users which do not create their own mdio device, but instead fetch it using a firmware description - namely the DPAA2 and FMAN_MEMAC drivers. In a previous patch set, lynx_pcs_create() was modified to increase the mdio device refcount, and lynx_pcs_destroy() to drop that refcount. The first two patches change these two drivers to put the reference which they hold immediately after lynx_pcs_create(), effectively handing the responsibility for maintaining the refcount to the Lynx PCS driver. A side effect of the first two patches is that lynx_get_mdio_device() is no longer used, so patch 3 removes it. Patch 4 adds a new helper - lynx_pcs_create_fwnode(), which creates a Lynx PCS instance from the fwnode. Patch 5 and 6 convert the two drivers to make use of this new helper, which simply has to find the mdio device, and then create the Lynx PCS from that. With those conversions done, lynx_pcs_create() is no longer required outside pcs-lynx.c, so remove it from public view. Patch 8 we changes lynx_pcs_create() to return an error-pointer rather than NULL to bring consistency to the return style, and means that we can remove the NULL-to-error-pointer conversion from both lynx_pcs_create_fwnode() and lynx_pcs_create_mdiodev(). Patch 9 adds a check for the fwnode being available, and returns an -ENODEV error pointer if unavailable. Patch 10 removes this check from DPAA2, detecting the error pointer value to continue printing the helpful message. Patch 11 removes this check from fman_memac, and in doing so fixes a bug where if the node is unavailable, the reference count is not dropped. ==================== Link: https://lore.kernel.org/r/ZIBwuw+IuGQo5yV8@shell.armlinux.org.ukSigned-off-by:
-
Russell King (Oracle) authored
Use pcs-lynx's check rather than our own when determining if the device is available. This fixes a bug where the reference gained by of_parse_phandle() is not dropped if the device is not available. Signed-off-by:
Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by:
-
Russell King (Oracle) authored
Use pcs-lynx's check rather than our own when determining if the device is available. Signed-off-by:
-
Russell King (Oracle) authored
Check that the fwnode is marked as available prior to trying to lookup the PCS device, and return -ENODEV if unavailable. Document the return codes from lynx_pcs_create_fwnode(). Signed-off-by:
-
Russell King (Oracle) authored
Change lynx_pcs_create() to return an error-pointer on failure to allocate memory, rather than returning NULL. This allows the removal of the conversion in lynx_pcs_create_fwnode() and lynx_pcs_create_mdiodev(). Signed-off-by:
-
Russell King (Oracle) authored
We no longer need to export lynx_pcs_create() for drivers to use as we now have all the functionality we need in the two new creation helpers. Remove the export and prototype, and make it static. Signed-off-by:
-
Russell King (Oracle) authored
Use lynx_pcs_create_fwnode() to create a lynx PCS from a fwnode handle. Signed-off-by:
-
Russell King (Oracle) authored
Use lynx_pcs_create_fwnode() to create a lynx PCS from a fwnode handle. Signed-off-by:
-
Russell King (Oracle) authored
Add a helper to create a lynx PCS from a fwnode handle. Signed-off-by:
-
Russell King (Oracle) authored
lynx_get_mdio_device() is no longer necessary, let's remove it so the lynx PCS code is always managing the lifetime of the mdiodev. Signed-off-by:
-
Russell King (Oracle) authored
Put the mdiodev after lynx_pcs_create() so that the Lynx PCS driver can manage the lifetime of the mdiodev its using. Signed-off-by:
-
Russell King (Oracle) authored
Put the mdiodev after lynx_pcs_create() so that the Lynx PCS driver can manage the lifetime of the mdiodev its using. Signed-off-by:
-
Jiaxun Yang authored
MIPS Boston board, which is using MIPS_GENERIC kernel is using EG20T PCH and thus need this driver. Dependency of PCH_GBE, PTP_1588_CLOCK_PCH is also fixed for MIPS_GENERIC. Note that CONFIG_PCH_GBE is selected in arch/mips/configs/generic/ board-boston.config for a while, some how it's never wired up in Kconfig. Signed-off-by:
Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by:
-
Ido Schimmel authored
The recently added 'VXLAN_F_LOCALBYPASS' flag is set by default on VXLAN devices and denotes a behavior that is irrelevant for the hardware data path. Add it to the lists of IPv4 and IPv6 supported flags to avoid rejecting offload of VXLAN devices which have this flag set. Fixes: 69474a8a ("net: vxlan: Add nolocalbypass option to vxlan.") Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Reviewed-by:
Petr Machata <petrm@nvidia.com> Signed-off-by:
-
- 08 Jun, 2023 12 commits
-
-
Jakub Kicinski authored
Jakub Kicinski says: ==================== tools: ynl: generate code for the devlink family Another chunk of changes to support more capabilities in the YNL code gen. Devlink brings in deep nesting and directional messages (requests and responses have different IDs). We need a healthy dose of codegen changes to support those (I wasn't planning to support code gen for "directional" families initially, but the importance of devlink and ethtool is undeniable). ==================== Link: https://lore.kernel.org/r/20230607202403.1089925-1-kuba@kernel.orgSigned-off-by:
-
Jakub Kicinski authored
Add a sample to show off how to issue basic devlink requests. For added testing issue get requests while walking a dump. $ ./devlink netdevsim/netdevsim1: driver: netdevsim running fw: fw.mgmt: 10.20.30 ... netdevsim/netdevsim2: driver: netdevsim running fw: fw.mgmt: 10.20.30 ... Signed-off-by: -
Jakub Kicinski authored
Admittedly the devlink.yaml spec is fairly limitted, it only covers basic device get and info-get ops. That's sufficient to be useful (monitoring FW versions in the fleet). Plus it gives us a chance to exercise deep nesting and directional messaging in YNL. Signed-off-by: -
Jakub Kicinski authored
Renegerate code after dropping forward declarations for policies. Signed-off-by: -
Jakub Kicinski authored
Now that all nested types have structs and are sorted topologically there should be no need to generate forward declarations for policies. Signed-off-by: -
Jakub Kicinski authored
So far we had only created structures for nested types nested directly in messages (second level of attrs so to speak). Walk types in depth to support deeper nesting. Signed-off-by: -
Jakub Kicinski authored
We only render parse and netlink generation helpers as needed, to avoid generating dead code. Propagate the information from first- and second-layer attribute sets onto all children. Otherwise devlink won't work, it has a lot more levels of nesting. Signed-off-by: -
Jakub Kicinski authored
We need to sort the structures to avoid the need for forward declarations. While at it remove the sort of structs when rendering, it doesn't do anything. Signed-off-by: -
Jakub Kicinski authored
I think that user space code gen for directional specs works after recent changes. Let them through. Signed-off-by: -
Jakub Kicinski authored
Move generating strmap lookup function to a helper. No functional changes. Signed-off-by: -
Jakub Kicinski authored
In preparation for supporting families which use different msg ids to and from the kernel - make sure the ids in op strmap are correct. The map is expected to be used mostly for notifications, don't generate a separate map for the "to kernel" direction. Signed-off-by: -
Jakub Kicinski authored
Python YNL is much more forgiving than the C code gen in terms of the spec completeness. Fill in a handful of devlink details to make the spec usable in C. Signed-off-by:
-
