- 09 Feb, 2017 8 commits
-
-
Alexander Stein authored
commit cdca06e4 upstream. According to VLI64 Intel Atom E3800 Specification Update (#329901) concurrent read accesses may result in returning 0xffffffff and write accesses may be dropped silently. To workaround all accesses must be protected by locks. Signed-off-by:
Alexander Stein <alexander.stein@systec-electronic.com> Acked-by:
Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by:
Linus Walleij <linus.walleij@linaro.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Johan Hovold authored
commit 8e9faa15 upstream. In case of a zero-length report, the gpio direction_input callback would currently return success instead of an errno. Fixes: 1ffb3c40 ("HID: cp2112: make transfer buffers DMA capable") Signed-off-by:
Johan Hovold <johan@kernel.org> Reviewed-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com> Signed-off-by:
Jiri Kosina <jkosina@suse.cz> Signed-off-by:
-
Johan Hovold authored
commit 7a7b5df8 upstream. A recent commit fixing DMA-buffers on stack added a shared transfer buffer protected by a spinlock. This is broken as the USB HID request callbacks can sleep. Fix this up by replacing the spinlock with a mutex. Fixes: 1ffb3c40 ("HID: cp2112: make transfer buffers DMA capable") Signed-off-by:
-
Max Filippov authored
commit 4b3e6f2e upstream. Commit bf15f86b ("xtensa: initialize MMU before jumping to reset vector") calls MMU management functions even when CONFIG_MMU is not selected. That breaks noMMU build on cores with MMU. Don't manage MMU when CONFIG_MMU is not selected. Signed-off-by:
Max Filippov <jcmvbkbc@gmail.com> Signed-off-by:
-
Ard Biesheuvel authored
commit c8f325a5 upstream. Some AArch64 UEFI implementations disable the MMU in ExitBootServices(), after which unaligned accesses to RAM are no longer supported. Commit: abfb7b68 ("efi/libstub/arm*: Pass latest memory map to the kernel") fixed an issue in the memory map handling of the stub FDT code, but inadvertently created an issue with such firmware, by moving some of the FDT manipulation to after the invocation of ExitBootServices(). Given that the stub's libfdt implementation uses the ordinary, accelerated string functions, which rely on hardware handling of unaligned accesses, manipulating the FDT with the MMU off may result in alignment faults. So fix the situation by moving the update_fdt_memmap() call into the callback function invoked by efi_exit_boot_services() right before it calls the ExitBootServices() UEFI service (which is arguably a better place for it anyway) Note that disabling the MMU in ExitBootServices() is not compliant with the UEFI spec, and carries great risk due to the fact that switching from cached to uncached memory accesses halfway through compiler generated code (i.e., involving a stack) can never be done in a way that is architecturally safe. Fixes: abfb7b68 ("efi/libstub/arm*: Pass latest memory map to the kernel") Signed-off-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> Tested-by:
Riku Voipio <riku.voipio@linaro.org> Cc: mark.rutland@arm.com Cc: linux-efi@vger.kernel.org Cc: matt@codeblueprint.co.uk Cc: leif.lindholm@linaro.org Cc: linux-arm-kernel@lists.infradead.org Link: http://lkml.kernel.org/r/1485971102-23330-2-git-send-email-ard.biesheuvel@linaro.orgSigned-off-by:
Ingo Molnar <mingo@kernel.org> Signed-off-by:
-
Jiri Kosina authored
commit bf29bddf upstream. Commit: 12976670 ("x86/efi: Only map RAM into EFI page tables if in mixed-mode") stopped creating 1:1 mappings for all RAM, when running in native 64-bit mode. It turns out though that there are 64-bit EFI implementations in the wild (this particular problem has been reported on a Lenovo Yoga 710-11IKB), which still make use of the first physical page for their own private use, even though they explicitly mark it EFI_CONVENTIONAL_MEMORY in the memory map. In case there is no mapping for this particular frame in the EFI pagetables, as soon as firmware tries to make use of it, a triple fault occurs and the system reboots (in case of the Yoga 710-11IKB this is very early during bootup). Fix that by always mapping the first page of physical memory into the EFI pagetables. We're free to hand this page to the BIOS, as trim_bios_range() will reserve the first page and isolate it away from memory allocators anyway. Note that just reverting 12976670 alone is not enough on v4.9-rc1+ to fix the regression on affected hardware, as this commit: ab72a27d ("x86/efi: Consolidate region mapping logic") later made the first physical frame not to be mapped anyway. Reported-by:
Hanka Pavlikova <hanka@ucw.cz> Signed-off-by:
Matt Fleming <matt@codeblueprint.co.uk> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Borislav Petkov <bp@suse.de> Cc: Laura Abbott <labbott@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vojtech Pavlik <vojtech@ucw.cz> Cc: Waiman Long <waiman.long@hpe.com> Cc: linux-efi@vger.kernel.org Fixes: 12976670 ("x86/efi: Only map RAM into EFI page tables if in mixed-mode") Link: http://lkml.kernel.org/r/20170127222552.22336-1-matt@codeblueprint.co.uk [ Tidied up the changelog and the comment. ] Signed-off-by:
-
Eryu Guan authored
commit 3a4b77cd upstream. Ralf Spenneberg reported that he hit a kernel crash when mounting a modified ext4 image. And it turns out that kernel crashed when calculating fs overhead (ext4_calculate_overhead()), this is because the image has very large s_first_meta_bg (debug code shows it's 842150400), and ext4 overruns the memory in count_overhead() when setting bitmap buffer, which is PAGE_SIZE. ext4_calculate_overhead(): buf = get_zeroed_page(GFP_NOFS); <=== PAGE_SIZE buffer blks = count_overhead(sb, i, buf); count_overhead(): for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { <=== j = 842150400 ext4_set_bit(EXT4_B2C(sbi, s++), buf); <=== buffer overrun count++; } This can be reproduced easily for me by this script: #!/bin/bash rm -f fs.img mkdir -p /mnt/ext4 fallocate -l 16M fs.img mke2fs -t ext4 -O bigalloc,meta_bg,^resize_inode -F fs.img debugfs -w -R "ssv first_meta_bg 842150400" fs.img mount -o loop fs.img /mnt/ext4 Fix it by validating s_first_meta_bg first at mount time, and refusing to mount if its value exceeds the largest possible meta_bg number. Reported-by:
Ralf Spenneberg <ralf@os-t.de> Signed-off-by:
Eryu Guan <guaneryu@gmail.com> Signed-off-by:
Theodore Ts'o <tytso@mit.edu> Reviewed-by:
Andreas Dilger <adilger@dilger.ca> Signed-off-by:
-
Bjorn Helgaas authored
commit 030305d6 upstream. In a struct pcie_link_state, link->root points to the pcie_link_state of the root of the PCIe hierarchy. For the topmost link, this points to itself (link->root = link). For others, we copy the pointer from the parent (link->root = link->parent->root). Previously we recognized that Root Ports originated PCIe hierarchies, but we treated PCI/PCI-X to PCIe Bridges as being in the middle of the hierarchy, and when we tried to copy the pointer from link->parent->root, there was no parent, and we dereferenced a NULL pointer: BUG: unable to handle kernel NULL pointer dereference at 0000000000000090 IP: [<ffffffff9e424350>] pcie_aspm_init_link_state+0x170/0x820 Recognize that PCI/PCI-X to PCIe Bridges originate PCIe hierarchies just like Root Ports do, so link->root for these devices should also point to itself. Fixes: 51ebfc92 ("PCI: Enumerate switches below PCI-to-PCIe bridges") Link: https://bugzilla.kernel.org/show_bug.cgi?id=193411 Link: https://bugzilla.opensuse.org/show_bug.cgi?id=1022181 Tested-by: lists@ssl-mail.com Tested-by:
Jayachandran C. <jnair@caviumnetworks.com> Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com> Signed-off-by:
-
- 04 Feb, 2017 32 commits
-
-
Greg Kroah-Hartman authored
-
Darrick J. Wong authored
commit c364b6d0 upstream. In a bmapx call, bmv_count is the total size of the array, including the zeroth element that userspace uses to supply the search key. The output array starts at offset 1 so that we can set up the user for the next invocation. Since we now can split an extent into multiple bmap records due to shared/unshared status, we have to be careful that we don't overflow the output array. In the original patch f86f4037 ("xfs: teach get_bmapx about shared extents and the CoW fork") I used cur_ext (the output index) to check for overflows, albeit with an off-by-one error. Since nexleft no longer describes the number of unfilled slots in the output, we can rip all that out and use cur_ext for the overflow check directly. Failure to do this causes heap corruption in bmapx callers such as xfs_io and xfs_scrub. xfs/328 can reproduce this problem. Reviewed-by:
Eric Sandeen <sandeen@redhat.com> Signed-off-by:
Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by:
-
Darrick J. Wong authored
commit 2aa6ba7b upstream. If we try to allocate memory pages to back an xfs_buf that we're trying to read, it's possible that we'll be so short on memory that the page allocation fails. For a blocking read we'll just wait, but for readahead we simply dump all the pages we've collected so far. Unfortunately, after dumping the pages we neglect to clear the _XBF_PAGES state, which means that the subsequent call to xfs_buf_free thinks that b_pages still points to pages we own. It then double-frees the b_pages pages. This results in screaming about negative page refcounts from the memory manager, which xfs oughtn't be triggering. To reproduce this case, mount a filesystem where the size of the inodes far outweighs the availalble memory (a ~500M inode filesystem on a VM with 300MB memory did the trick here) and run bulkstat in parallel with other memory eating processes to put a huge load on the system. The "check summary" phase of xfs_scrub also works for this purpose. Signed-off-by:
-
Christoph Hellwig authored
commit 493611eb upstream. With COW files they are the hotpath, just like for files with the extent size hint attribute. We really shouldn't micro-manage anything but failure cases with unlikely. Additionally Arnd Bergmann recently reported that one of these two unlikely annotations causes link failures together with an upcoming kernel instrumentation patch, so let's get rid of it ASAP. Signed-off-by:
Christoph Hellwig <hch@lst.de> Reported-by:
Arnd Bergmann <arnd@arndb.de> Reviewed-by:
-
Brian Foster authored
commit 5a93790d upstream. xfs_attr_[get|remove]() have unlocked attribute fork checks to optimize away a lock cycle in cases where the fork does not exist or is otherwise empty. This check is not safe, however, because an attribute fork short form to extent format conversion includes a transient state that causes the xfs_inode_hasattr() check to fail. Specifically, xfs_attr_shortform_to_leaf() creates an empty extent format attribute fork and then adds the existing shortform attributes to it. This means that lookup of an existing xattr can spuriously return -ENOATTR when racing against a setxattr that causes the associated format conversion. This was originally reproduced by an untar on a particularly configured glusterfs volume, but can also be reproduced on demand with properly crafted xattr requests. The format conversion occurs under the exclusive ilock. xfs_attr_get() and xfs_attr_remove() already have the proper locking and checks further down in the functions to handle this situation correctly. Drop the unlocked checks to avoid the spurious failure and rely on the existing logic. Signed-off-by:
Brian Foster <bfoster@redhat.com> Reviewed-by:
-
Darrick J. Wong authored
commit 83d230eb upstream. sb_dirblklog is added to sb_blocklog to compute the directory block size in bytes. Therefore, we must compare the sum of both those values against XFS_MAX_BLOCKSIZE_LOG, not just dirblklog. Signed-off-by:
-
Christoph Hellwig authored
commit d2b3964a upstream. Due to the way how xfs_iomap_write_allocate tries to convert the whole found extents from delalloc to real space we can run into a race condition with multiple threads doing writes to this same extent. For the non-COW case that is harmless as the only thing that can happen is that we call xfs_bmapi_write on an extent that has already been converted to a real allocation. For COW writes where we move the extent from the COW to the data fork after I/O completion the race is, however, not quite as harmless. In the worst case we are now calling xfs_bmapi_write on a region that contains hole in the COW work, which will trip up an assert in debug builds or lead to file system corruption in non-debug builds. This seems to be reproducible with workloads of small O_DSYNC write, although so far I've not managed to come up with a with an isolated reproducer. The fix for the issue is relatively simple: tell xfs_bmapi_write that we are only asked to convert delayed allocations and skip holes in that case. Signed-off-by:
-
Arnd Bergmann authored
commit fd29f7af upstream. A harmless warning just got introduced: fs/xfs/libxfs/xfs_dir2.h:40:8: error: type qualifiers ignored on function return type [-Werror=ignored-qualifiers] Removing the 'const' modifier avoids the warning and has no other effect. Fixes: 1fc4d33f ("xfs: replace xfs_mode_to_ftype table with switch statement") Signed-off-by:
-
Eric Sandeen authored
commit 657bdfb7 upstream. The GETNEXTQOTA ioctl takes whatever ID is sent in, and looks for the next active quota for an user equal or higher to that ID. But if we are at the maximum ID and then ask for the "next" one, we may wrap back to zero. In this case, userspace may loop forever, because it will start querying again at zero. We'll fix this in userspace as well, but for the kernel, return -ENOENT if we ask for the next quota ID past UINT_MAX so the caller knows to stop. Signed-off-by:
-
Amir Goldstein authored
commit a324cbf1 upstream. Check for invalid file type in xfs_dinode_verify() and fail to load the inode structure from disk. Reviewed-by:
Amir Goldstein <amir73il@gmail.com> Reviewed-by:
-
Amir Goldstein authored
commit fab8eef8 upstream. The helper xfs_dentry_to_name() is used by 2 different classes of callers: Callers that pass zero mode and don't care about the returned name.type field and Callers that pass non zero mode and do care about the name.type field. Change xfs_dentry_to_name() to not take the mode argument and change the call sites of the first class to not pass the mode argument. Create a new helper xfs_dentry_mode_to_name() which does pass the mode argument and returns -EFSCORRUPTED if mode is invalid. Callers that translate non zero mode to on-disk file type now check the return value and will export the error to user instead of staging an invalid file type to be written to directory entry. Signed-off-by:
-
Amir Goldstein authored
commit 1fc4d33f. The size of the xfs_mode_to_ftype[] conversion table was too small to handle an invalid value of mode=S_IFMT. Instead of fixing the table size, replace the conversion table with a conversion helper that uses a switch statement. Suggested-by:
-
Amir Goldstein authored
commit b597dd53 upstream. xfs_dir2.h dereferences some data types in inline functions and fails to include those type definitions, e.g.: xfs_dir2_data_aoff_t, struct xfs_da_geometry. Signed-off-by:
-
Amir Goldstein authored
commit 3c6f46ea upstream. This changes fixes an assertion hit when fuzzing on-disk i_mode values. The easy case to fix is when changing an empty file i_mode to S_IFDIR. In this case, xfs_dinode_verify() detects an illegal zero size for directory and fails to load the inode structure from disk. For the case of non empty file whose i_mode is changed to S_IFDIR, the ASSERT() statement in xfs_dir2_isblock() is replaced with return -EFSCORRUPTED, to avoid interacting with corrupted jusk also when XFS_DEBUG is disabled. Suggested-by:
-
Amir Goldstein authored
commit bf46ecc3 upstream. The ASSERT() condition is the normal case, not the exception, so testing the condition should be likely(), not unlikely(). Reviewed-by:
-
Christoph Hellwig authored
commit 84a4620c upstream. There are only two reasons for xfs_log_force / xfs_log_force_lsn to fail: one is an I/O error, for which xlog_bdstrat already logs a warning, and the second is an already shutdown log due to a previous I/O errors. In the latter case we'll already have a previous indication for the actual error, but the large stream of misleading warnings from xfs_log_force will probably scroll it out of the message buffer. Simply removing the warnings thus makes the XFS log reporting significantly better. Signed-off-by:
Carlos Maiolino <cmaiolino@redhat.com> Signed-off-by:
-
Christoph Hellwig authored
commit 12ef8301 upstream. ->total is a bit of an odd parameter passed down to the low-level allocator all the way from the high-level callers. It's supposed to contain the maximum number of blocks to be allocated for the whole transaction [1]. But in xfs_iomap_write_allocate we only convert existing delayed allocations and thus only have a minimal block reservation for the current transaction, so xfs_alloc_space_available can't use it for the allocation decisions. Use the maximum of args->total and the calculated block requirement to make a decision. We probably should get rid of args->total eventually and instead apply ->minleft more broadly, but that will require some extensive changes all over. [1] which creates lots of confusion as most callers don't decrement it once doing a first allocation. But that's for a separate series. Signed-off-by:
-
Christoph Hellwig authored
commit 54fee133 upstream. We must decide in xfs_alloc_fix_freelist if we can perform an allocation from a given AG is possible or not based on the available space, and should not fail the allocation past that point on a healthy file system. But currently we have two additional places that second-guess xfs_alloc_fix_freelist: xfs_alloc_ag_vextent tries to adjust the maxlen parameter to remove the reservation before doing the allocation (but ignores the various minium freespace requirements), and xfs_alloc_fix_minleft tries to fix up the allocated length after we've found an extent, but ignores the reservations and also doesn't take the AGFL into account (and thus fails allocations for not matching minlen in some cases). Remove all these later fixups and just correct the maxlen argument inside xfs_alloc_fix_freelist once we have the AGF buffer locked. Signed-off-by:
-
Christoph Hellwig authored
commit 255c5162 upstream. We can't just set minleft to 0 when we're low on space - that's exactly what we need minleft for: to protect space in the AG for btree block allocations when we are low on free space. Signed-off-by:
-
Christoph Hellwig authored
commit 5149fd32 upstream. Setting aside 4 blocks globally for bmbt splits isn't all that useful, as different threads can allocate space in parallel. Bump it to 4 blocks per AG to allow each thread that is currently doing an allocation to dip into it separately. Without that we may no have enough reserved blocks if there are enough parallel transactions in an almost out space file system that all run into bmap btree splits. Signed-off-by:
-
Florian Fainelli authored
[ Upstream commit f154be24 ] Commit 448b4482 ("net: dsa: Add lockdep class to tx queues to avoid lockdep splat") removed the netif_device_detach() call done in dsa_slave_suspend() which is necessary, and paired with a corresponding netif_device_attach(), bring it back. Fixes: 448b4482 ("net: dsa: Add lockdep class to tx queues to avoid lockdep splat") Signed-off-by:
Florian Fainelli <f.fainelli@gmail.com> Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
David S. Miller <davem@davemloft.net> Signed-off-by:
-
Robert Shearman authored
[ Upstream commit 85c81401 ] When attempting to free lwtunnel state after the module for the encap has been unloaded an oops occurs: BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 IP: lwtstate_free+0x18/0x40 [..] task: ffff88003e372380 task.stack: ffffc900001fc000 RIP: 0010:lwtstate_free+0x18/0x40 RSP: 0018:ffff88003fd83e88 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff88002bbb3380 RCX: ffff88000c91a300 [..] Call Trace: <IRQ> free_fib_info_rcu+0x195/0x1a0 ? rt_fibinfo_free+0x50/0x50 rcu_process_callbacks+0x2d3/0x850 ? rcu_process_callbacks+0x296/0x850 __do_softirq+0xe4/0x4cb irq_exit+0xb0/0xc0 smp_apic_timer_interrupt+0x3d/0x50 apic_timer_interrupt+0x93/0xa0 [..] Code: e8 6e c6 fc ff 89 d8 5b 5d c3 bb de ff ff ff eb f4 66 90 66 66 66 66 90 55 48 89 e5 53 0f b7 07 48 89 fb 48 8b 04 c5 00 81 d5 81 <48> 8b 40 08 48 85 c0 74 13 ff d0 48 8d 7b 20 be 20 00 00 00 e8 The problem is after the module for the encap can be unloaded the corresponding ops is removed and is thus NULL here. Modules implementing lwtunnel ops should not be allowed to unload while there is state alive using those ops, so grab the module reference for the ops on creating lwtunnel state and of course release the reference when freeing the state. Fixes: 1104d9ba ("lwtunnel: Add destroy state operation") Signed-off-by:
Robert Shearman <rshearma@brocade.com> Signed-off-by:
-
Robert Shearman authored
[ Upstream commit 88ff7334 ] Modules implementing lwtunnel ops should not be allowed to unload while there is state alive using those ops, so specify the owning module for all lwtunnel ops. Signed-off-by:
-
Bjørn Mork authored
[ Upstream commit 5b9f5751 ] Another rebranded Novatel E371. qmi_wwan should drive this device, while cdc_ether should ignore it. Even though the USB descriptors are plain CDC-ETHER that USB interface is a QMI interface. Ref commit 7fdb7846 ("qmi_wwan/cdc_ether: add device IDs for Dell 5804 (Novatel E371) WWAN card") Cc: Dan Williams <dcbw@redhat.com> Signed-off-by:
Bjørn Mork <bjorn@mork.no> Signed-off-by:
-
WANG Cong authored
[ Upstream commit 0fb44559 ] Dmitry reported a deadlock scenario: unix_bind() path: u->bindlock ==> sb_writer do_splice() path: sb_writer ==> pipe->mutex ==> u->bindlock In the unix_bind() code path, unix_mknod() does not have to be done with u->bindlock held, since it is a pure fs operation, so we can just move unix_mknod() out. Reported-by:
Dmitry Vyukov <dvyukov@google.com> Tested-by:
Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by:
-
hayeswang authored
[ Upstream commit 6a0b76c0 ] Runtime suspend shouldn't be executed if the tx queue is not empty, because the device is not idle. Signed-off-by:
Hayes Wang <hayeswang@realtek.com> Signed-off-by:
-
David Ahern authored
[ Upstream commit 9f427a0e ] MPLS multipath for LSR is broken -- always selecting the first nexthop in the one label case. For example: $ ip -f mpls ro ls 100 nexthop as to 200 via inet 172.16.2.2 dev virt12 nexthop as to 300 via inet 172.16.3.2 dev virt13 101 nexthop as to 201 via inet6 2000:2::2 dev virt12 nexthop as to 301 via inet6 2000:3::2 dev virt13 In this example incoming packets have a single MPLS labels which means BOS bit is set. The BOS bit is passed from mpls_forward down to mpls_multipath_hash which never processes the hash loop because BOS is 1. Update mpls_multipath_hash to process the entire label stack. mpls_hdr_len tracks the total mpls header length on each pass (on pass N mpls_hdr_len is N * sizeof(mpls_shim_hdr)). When the label is found with the BOS set it verifies the skb has sufficient header for ipv4 or ipv6, and find the IPv4 and IPv6 header by using the last mpls_hdr pointer and adding 1 to advance past it. With these changes I have verified the code correctly sees the label, BOS, IPv4 and IPv6 addresses in the network header and icmp/tcp/udp traffic for ipv4 and ipv6 are distributed across the nexthops. Fixes: 1c78efa8 ("mpls: flow-based multipath selection") Acked-by:
David Ahern <dsa@cumulusnetworks.com> Signed-off-by:
-
Ivan Vecera authored
[ Upstream commit b6677449 ] Any bridge options specified during link creation (e.g. ip link add) are ignored as br_dev_newlink() does not process them. Use br_changelink() to do it. Fixes: 13323516 ("bridge: implement rtnl_link_ops->changelink") Signed-off-by:
Ivan Vecera <cera@cera.cz> Reviewed-by:
Jiri Pirko <jiri@mellanox.com> Signed-off-by:
-
Eric Dumazet authored
[ Upstream commit e048fc50 ] A driver using dev_alloc_page() must not reuse a page allocated from emergency memory reserve. Otherwise all packets using this page will be immediately dropped, unless for very specific sockets having SOCK_MEMALLOC bit set. This issue might be hard to debug, because only a fraction of received packets would be dropped. Fixes: 4415a031 ("net/mlx5e: Implement RX mapped page cache for page recycle") Signed-off-by:
Eric Dumazet <edumazet@google.com> Cc: Tariq Toukan <tariqt@mellanox.com> Cc: Saeed Mahameed <saeedm@mellanox.com> Acked-by:
Saeed Mahameed <saeedm@mellanox.com> Signed-off-by:
-
Alexey Kodanev authored
[ Upstream commit 0dbd7ff3 ] Found that if we run LTP netstress test with large MSS (65K), the first attempt from server to send data comparable to this MSS on fastopen connection will be delayed by the probe timer. Here is an example: < S seq 0:0 win 43690 options [mss 65495 wscale 7 tfo cookie] length 32 > S. seq 0:0 ack 1 win 43690 options [mss 65495 wscale 7] length 0 < . ack 1 win 342 length 0 Inside tcp_sendmsg(), tcp_send_mss() returns max MSS in 'mss_now', as well as in 'size_goal'. This results the segment not queued for transmition until all the data copied from user buffer. Then, inside __tcp_push_pending_frames(), it breaks on send window test and continues with the check probe timer. Fragmentation occurs in tcp_write_wakeup()... +0.2 > P. seq 1:43777 ack 1 win 342 length 43776 < . ack 43777, win 1365 length 0 > P. seq 43777:65001 ack 1 win 342 options [...] length 21224 ... This also contradicts with the fact that we should bound to the half of the window if it is large. Fix this flaw by correctly initializing max_window. Before that, it could have large values that affect further calculations of 'size_goal'. Fixes: 168a8f58 ("tcp: TCP Fast Open Server - main code path") Signed-off-by:
Alexey Kodanev <alexey.kodanev@oracle.com> Acked-by:
-
Kefeng Wang authored
[ Upstream commit 03e4deff ] Just like commit 4acd4945 ("ipv6: addrconf: Avoid calling netdevice notifiers with RCU read-side lock"), it is unnecessary to make addrconf_disable_change() use RCU iteration over the netdev list, since it already holds the RTNL lock, or we may meet Illegal context switch in RCU read-side critical section. Signed-off-by:
Kefeng Wang <wangkefeng.wang@huawei.com> Signed-off-by:
-
David Ahern authored
[ Upstream commit 9ed59592] Trying to add an mpls encap route when the MPLS modules are not loaded hangs. For example: CONFIG_MPLS=y CONFIG_NET_MPLS_GSO=m CONFIG_MPLS_ROUTING=m CONFIG_MPLS_IPTUNNEL=m $ ip route add 10.10.10.10/32 encap mpls 100 via inet 10.100.1.2 The ip command hangs: root 880 826 0 21:25 pts/0 00:00:00 ip route add 10.10.10.10/32 encap mpls 100 via inet 10.100.1.2 $ cat /proc/880/stack [<ffffffff81065a9b>] call_usermodehelper_exec+0xd6/0x134 [<ffffffff81065efc>] __request_module+0x27b/0x30a [<ffffffff814542f6>] lwtunnel_build_state+0xe4/0x178 [<ffffffff814aa1e4>] fib_create_info+0x47f/0xdd4 [<ffffffff814ae451>] fib_table_insert+0x90/0x41f [<ffffffff814a8010>] inet_rtm_newroute+0x4b/0x52 ... modprobe is trying to load rtnl-lwt-MPLS: root 881 5 0 21:25 ? 00:00:00 /sbin/modprobe -q -- rtnl-lwt-MPLS and it hangs after loading mpls_router: $ cat /proc/881/stack [<ffffffff81441537>] rtnl_lock+0x12/0x14 [<ffffffff8142ca2a>] register_netdevice_notifier+0x16/0x179 [<ffffffffa0033025>] mpls_init+0x25/0x1000 [mpls_router] [<ffffffff81000471>] do_one_initcall+0x8e/0x13f [<ffffffff81119961>] do_init_module+0x5a/0x1e5 [<ffffffff810bd070>] load_module+0x13bd/0x17d6 ... The problem is that lwtunnel_build_state is called with rtnl lock held preventing mpls_init from registering. Given the potential references held by the time lwtunnel_build_state it can not drop the rtnl lock to the load module. So, extract the module loading code from lwtunnel_build_state into a new function to validate the encap type. The new function is called while converting the user request into a fib_config which is well before any table, device or fib entries are examined. Fixes: 745041e2 ("lwtunnel: autoload of lwt modules") Signed-off-by:
-
