November 2018 - Linux-stable-mirror

[PATCH] PCI: vmd: White list for fast interrupt handlers

by Heitke, Kenneth

Please apply the follow patch from kernel 4.18 to kernel 4.14 Commit: a7f58b9ecfd3c0f63703ec10f4a592cc38dbd1b8 This patch fixes a performance issue with some devices being accessed using the Volume Management Device (vmd). I have included an updated copy of the patch because the parent directory of the file has changed between the 4.14 and 4.18 releases 4.14: drivers/pci/host/vmd.c 4.18: drivers/pci/controller/vmd.c Thank you, Kenneth Heitke <Kenneth.heitke(a)intel.com> --- From: Keith Busch <keith.busch(a)intel.com> Date: Tue, 8 May 2018 10:00:22 -0600 Subject: [PATCH] PCI: vmd: White list for fast interrupt handlers Devices with slow interrupt handlers are significantly harming performance when their interrupt vector is shared with a fast device. Create a class code white list for devices with known fast interrupt handlers and let all other devices share a single vector so that they don't interfere with performance. At the moment, only the NVM Express class code is on the list, but more may be added if VMD users desire to use other low-latency devices in these domains. Signed-off-by: Keith Busch <keith.busch(a)intel.com> [lorenzo.pieralisi(a)arm.com: changelog] Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi(a)arm.com> Acked-by: Jon Derrick: <jonathan.derrick(a)intel.com> --- drivers/pci/host/vmd.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/drivers/pci/host/vmd.c b/drivers/pci/host/vmd.c index 509893b..2537b02 100644 --- a/drivers/pci/host/vmd.c +++ b/drivers/pci/host/vmd.c @@ -183,9 +183,20 @@ static struct vmd_irq_list *vmd_next_irq(struct vmd_dev *vmd, struct msi_desc *d int i, best = 1; unsigned long flags; - if (pci_is_bridge(msi_desc_to_pci_dev(desc)) || vmd->msix_count == 1) + if (vmd->msix_count == 1) return &vmd->irqs[0]; + /* + * White list for fast-interrupt handlers. All others will share the + * "slow" interrupt vector. + */ + switch (msi_desc_to_pci_dev(desc)->class) { + case PCI_CLASS_STORAGE_EXPRESS: + break; + default: + return &vmd->irqs[0]; + } + raw_spin_lock_irqsave(&list_lock, flags); for (i = 1; i < vmd->msix_count; i++) if (vmd->irqs[i].count < vmd->irqs[best].count) -- 2.7.4

6 years, 2 months

2
1
0 0

Patch to consider for stable 3.18, 4.4 and 4.9

by Tomasz Kramkowski

In September last year, Ben Hutchings submitted commit [9547837bdccb] for 3.16.48-rc1 and I informed him that it would be useless without [3f3752705dbd] (and that maybe [c3883fe06488] would be useful as well). Ben dropped the patch but suggested I email this list with the information of the other two patches but I never quite got around to it. Now I see Sasha Levin is submitting [3f3752705dbd] and [c3883fe06488] for 4.9, 4.4 and 3.18 it would now make sense to include [9547837bdccb]. This patch fixes a minor problem where a certain USB adapter for Sega Genesis controllers appears as one input device when it has two ports for two controllers. I imagine some users of emulator distributions might use stable kernels and might benefit from this fix. I'm actually not entirely sure that patch is something suitable for stable but since it was already submitted once then I don't think it hurts to bring it up again (despite it breaking stable-kernel-rules as far as I understand it). Commits mentioned: [9547837bdccb]: HID: usbhid: add quirk for innomedia INNEX GENESIS/ATARI adapter [3f3752705dbd]: HID: reject input outside logical range only if null state is set [c3883fe06488]: HID: clamp input to logical range if no null state If the patch [9547837bdccb] is not relevant then feel free to ignore this email. Thanks, -- Tomasz Kramkowski

6 years, 2 months

3
2
0 0

FAILED: patch "[PATCH] PCI/ASPM: Fix link_state teardown on device removal" failed to apply to 3.18-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 3.18-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From aeae4f3e5c38d47bdaef50446dc0ec857307df68 Mon Sep 17 00:00:00 2001 From: Lukas Wunner <lukas(a)wunner.de> Date: Tue, 4 Sep 2018 12:34:18 -0500 Subject: [PATCH] PCI/ASPM: Fix link_state teardown on device removal Upon removal of the last device on a bus, the link_state of the bridge leading to that bus is sought to be torn down by having pci_stop_dev() call pcie_aspm_exit_link_state(). When ASPM was originally introduced by commit 7d715a6c1ae5 ("PCI: add PCI Express ASPM support"), it determined whether the device being removed is the last one by calling list_empty() on the bridge's subordinate devices list. That didn't work because the device is only removed from the list slightly later in pci_destroy_dev(). Commit 3419c75e15f8 ("PCI: properly clean up ASPM link state on device remove") attempted to fix it by calling list_is_last(), but that's not correct either because it checks whether the device is at the *end* of the list, not whether it's the last one *left* in the list. If the user removes the device which happens to be at the end of the list via sysfs but other devices are preceding the device in the list, the link_state is torn down prematurely. The real fix is to move the invocation of pcie_aspm_exit_link_state() to pci_destroy_dev() and reinstate the call to list_empty(). Remove a duplicate check for dev->bus->self because pcie_aspm_exit_link_state() already contains an identical check. Fixes: 7d715a6c1ae5 ("PCI: add PCI Express ASPM support") Signed-off-by: Lukas Wunner <lukas(a)wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas(a)google.com> Cc: Shaohua Li <shaohua.li(a)intel.com> Cc: stable(a)vger.kernel.org # v2.6.26 diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c index 5326916715d2..f78860ce884b 100644 --- a/drivers/pci/pcie/aspm.c +++ b/drivers/pci/pcie/aspm.c @@ -991,7 +991,7 @@ void pcie_aspm_exit_link_state(struct pci_dev *pdev) * All PCIe functions are in one slot, remove one function will remove * the whole slot, so just wait until we are the last function left. */ - if (!list_is_last(&pdev->bus_list, &parent->subordinate->devices)) + if (!list_empty(&parent->subordinate->devices)) goto out; link = parent->link_state; diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c index 461e7fd2756f..e9c6b120cf45 100644 --- a/drivers/pci/remove.c +++ b/drivers/pci/remove.c @@ -25,9 +25,6 @@ static void pci_stop_dev(struct pci_dev *dev) pci_dev_assign_added(dev, false); } - - if (dev->bus->self) - pcie_aspm_exit_link_state(dev); } static void pci_destroy_dev(struct pci_dev *dev) @@ -41,6 +38,7 @@ static void pci_destroy_dev(struct pci_dev *dev) list_del(&dev->bus_list); up_write(&pci_bus_sem); + pcie_aspm_exit_link_state(dev); pci_bridge_d3_update(dev); pci_free_resources(dev); put_device(&dev->dev);

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] PCI/ASPM: Fix link_state teardown on device removal" failed to apply to 4.4-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.4-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From aeae4f3e5c38d47bdaef50446dc0ec857307df68 Mon Sep 17 00:00:00 2001 From: Lukas Wunner <lukas(a)wunner.de> Date: Tue, 4 Sep 2018 12:34:18 -0500 Subject: [PATCH] PCI/ASPM: Fix link_state teardown on device removal Upon removal of the last device on a bus, the link_state of the bridge leading to that bus is sought to be torn down by having pci_stop_dev() call pcie_aspm_exit_link_state(). When ASPM was originally introduced by commit 7d715a6c1ae5 ("PCI: add PCI Express ASPM support"), it determined whether the device being removed is the last one by calling list_empty() on the bridge's subordinate devices list. That didn't work because the device is only removed from the list slightly later in pci_destroy_dev(). Commit 3419c75e15f8 ("PCI: properly clean up ASPM link state on device remove") attempted to fix it by calling list_is_last(), but that's not correct either because it checks whether the device is at the *end* of the list, not whether it's the last one *left* in the list. If the user removes the device which happens to be at the end of the list via sysfs but other devices are preceding the device in the list, the link_state is torn down prematurely. The real fix is to move the invocation of pcie_aspm_exit_link_state() to pci_destroy_dev() and reinstate the call to list_empty(). Remove a duplicate check for dev->bus->self because pcie_aspm_exit_link_state() already contains an identical check. Fixes: 7d715a6c1ae5 ("PCI: add PCI Express ASPM support") Signed-off-by: Lukas Wunner <lukas(a)wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas(a)google.com> Cc: Shaohua Li <shaohua.li(a)intel.com> Cc: stable(a)vger.kernel.org # v2.6.26 diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c index 5326916715d2..f78860ce884b 100644 --- a/drivers/pci/pcie/aspm.c +++ b/drivers/pci/pcie/aspm.c @@ -991,7 +991,7 @@ void pcie_aspm_exit_link_state(struct pci_dev *pdev) * All PCIe functions are in one slot, remove one function will remove * the whole slot, so just wait until we are the last function left. */ - if (!list_is_last(&pdev->bus_list, &parent->subordinate->devices)) + if (!list_empty(&parent->subordinate->devices)) goto out; link = parent->link_state; diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c index 461e7fd2756f..e9c6b120cf45 100644 --- a/drivers/pci/remove.c +++ b/drivers/pci/remove.c @@ -25,9 +25,6 @@ static void pci_stop_dev(struct pci_dev *dev) pci_dev_assign_added(dev, false); } - - if (dev->bus->self) - pcie_aspm_exit_link_state(dev); } static void pci_destroy_dev(struct pci_dev *dev) @@ -41,6 +38,7 @@ static void pci_destroy_dev(struct pci_dev *dev) list_del(&dev->bus_list); up_write(&pci_bus_sem); + pcie_aspm_exit_link_state(dev); pci_bridge_d3_update(dev); pci_free_resources(dev); put_device(&dev->dev);

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] PCI/ASPM: Fix link_state teardown on device removal" failed to apply to 4.9-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.9-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From aeae4f3e5c38d47bdaef50446dc0ec857307df68 Mon Sep 17 00:00:00 2001 From: Lukas Wunner <lukas(a)wunner.de> Date: Tue, 4 Sep 2018 12:34:18 -0500 Subject: [PATCH] PCI/ASPM: Fix link_state teardown on device removal Upon removal of the last device on a bus, the link_state of the bridge leading to that bus is sought to be torn down by having pci_stop_dev() call pcie_aspm_exit_link_state(). When ASPM was originally introduced by commit 7d715a6c1ae5 ("PCI: add PCI Express ASPM support"), it determined whether the device being removed is the last one by calling list_empty() on the bridge's subordinate devices list. That didn't work because the device is only removed from the list slightly later in pci_destroy_dev(). Commit 3419c75e15f8 ("PCI: properly clean up ASPM link state on device remove") attempted to fix it by calling list_is_last(), but that's not correct either because it checks whether the device is at the *end* of the list, not whether it's the last one *left* in the list. If the user removes the device which happens to be at the end of the list via sysfs but other devices are preceding the device in the list, the link_state is torn down prematurely. The real fix is to move the invocation of pcie_aspm_exit_link_state() to pci_destroy_dev() and reinstate the call to list_empty(). Remove a duplicate check for dev->bus->self because pcie_aspm_exit_link_state() already contains an identical check. Fixes: 7d715a6c1ae5 ("PCI: add PCI Express ASPM support") Signed-off-by: Lukas Wunner <lukas(a)wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas(a)google.com> Cc: Shaohua Li <shaohua.li(a)intel.com> Cc: stable(a)vger.kernel.org # v2.6.26 diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c index 5326916715d2..f78860ce884b 100644 --- a/drivers/pci/pcie/aspm.c +++ b/drivers/pci/pcie/aspm.c @@ -991,7 +991,7 @@ void pcie_aspm_exit_link_state(struct pci_dev *pdev) * All PCIe functions are in one slot, remove one function will remove * the whole slot, so just wait until we are the last function left. */ - if (!list_is_last(&pdev->bus_list, &parent->subordinate->devices)) + if (!list_empty(&parent->subordinate->devices)) goto out; link = parent->link_state; diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c index 461e7fd2756f..e9c6b120cf45 100644 --- a/drivers/pci/remove.c +++ b/drivers/pci/remove.c @@ -25,9 +25,6 @@ static void pci_stop_dev(struct pci_dev *dev) pci_dev_assign_added(dev, false); } - - if (dev->bus->self) - pcie_aspm_exit_link_state(dev); } static void pci_destroy_dev(struct pci_dev *dev) @@ -41,6 +38,7 @@ static void pci_destroy_dev(struct pci_dev *dev) list_del(&dev->bus_list); up_write(&pci_bus_sem); + pcie_aspm_exit_link_state(dev); pci_bridge_d3_update(dev); pci_free_resources(dev); put_device(&dev->dev);

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] ext4: fix use-after-free race in ext4_remount()'s error path" failed to apply to 3.18-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 3.18-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 33458eaba4dfe778a426df6a19b7aad2ff9f7eec Mon Sep 17 00:00:00 2001 From: Theodore Ts'o <tytso(a)mit.edu> Date: Fri, 12 Oct 2018 09:28:09 -0400 Subject: [PATCH] ext4: fix use-after-free race in ext4_remount()'s error path It's possible for ext4_show_quota_options() to try reading s_qf_names[i] while it is being modified by ext4_remount() --- most notably, in ext4_remount's error path when the original values of the quota file name gets restored. Reported-by: syzbot+a2872d6feea6918008a9(a)syzkaller.appspotmail.com Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Cc: stable(a)kernel.org # 3.2+ diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 86e1bacac757..12f90d48ba61 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -1405,7 +1405,8 @@ struct ext4_sb_info { u32 s_min_batch_time; struct block_device *journal_bdev; #ifdef CONFIG_QUOTA - char *s_qf_names[EXT4_MAXQUOTAS]; /* Names of quota files with journalled quota */ + /* Names of quota files with journalled quota */ + char __rcu *s_qf_names[EXT4_MAXQUOTAS]; int s_jquota_fmt; /* Format of quota to use */ #endif unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */ diff --git a/fs/ext4/super.c b/fs/ext4/super.c index faf293ed8060..a221f1cdf704 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -914,6 +914,18 @@ static inline void ext4_quota_off_umount(struct super_block *sb) for (type = 0; type < EXT4_MAXQUOTAS; type++) ext4_quota_off(sb, type); } + +/* + * This is a helper function which is used in the mount/remount + * codepaths (which holds s_umount) to fetch the quota file name. + */ +static inline char *get_qf_name(struct super_block *sb, + struct ext4_sb_info *sbi, + int type) +{ + return rcu_dereference_protected(sbi->s_qf_names[type], + lockdep_is_held(&sb->s_umount)); +} #else static inline void ext4_quota_off_umount(struct super_block *sb) { @@ -965,7 +977,7 @@ static void ext4_put_super(struct super_block *sb) percpu_free_rwsem(&sbi->s_journal_flag_rwsem); #ifdef CONFIG_QUOTA for (i = 0; i < EXT4_MAXQUOTAS; i++) - kfree(sbi->s_qf_names[i]); + kfree(get_qf_name(sb, sbi, i)); #endif /* Debugging code just in case the in-memory inode orphan list @@ -1531,11 +1543,10 @@ static const char deprecated_msg[] = static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) { struct ext4_sb_info *sbi = EXT4_SB(sb); - char *qname; + char *qname, *old_qname = get_qf_name(sb, sbi, qtype); int ret = -1; - if (sb_any_quota_loaded(sb) && - !sbi->s_qf_names[qtype]) { + if (sb_any_quota_loaded(sb) && !old_qname) { ext4_msg(sb, KERN_ERR, "Cannot change journaled " "quota options when quota turned on"); @@ -1552,8 +1563,8 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) "Not enough memory for storing quotafile name"); return -1; } - if (sbi->s_qf_names[qtype]) { - if (strcmp(sbi->s_qf_names[qtype], qname) == 0) + if (old_qname) { + if (strcmp(old_qname, qname) == 0) ret = 1; else ext4_msg(sb, KERN_ERR, @@ -1566,7 +1577,7 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) "quotafile must be on filesystem root"); goto errout; } - sbi->s_qf_names[qtype] = qname; + rcu_assign_pointer(sbi->s_qf_names[qtype], qname); set_opt(sb, QUOTA); return 1; errout: @@ -1578,15 +1589,16 @@ static int clear_qf_name(struct super_block *sb, int qtype) { struct ext4_sb_info *sbi = EXT4_SB(sb); + char *old_qname = get_qf_name(sb, sbi, qtype); - if (sb_any_quota_loaded(sb) && - sbi->s_qf_names[qtype]) { + if (sb_any_quota_loaded(sb) && old_qname) { ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" " when quota turned on"); return -1; } - kfree(sbi->s_qf_names[qtype]); - sbi->s_qf_names[qtype] = NULL; + rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); + synchronize_rcu(); + kfree(old_qname); return 1; } #endif @@ -1961,7 +1973,7 @@ static int parse_options(char *options, struct super_block *sb, int is_remount) { struct ext4_sb_info *sbi = EXT4_SB(sb); - char *p; + char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; substring_t args[MAX_OPT_ARGS]; int token; @@ -1992,11 +2004,13 @@ static int parse_options(char *options, struct super_block *sb, "Cannot enable project quota enforcement."); return 0; } - if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { - if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) + usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); + grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); + if (usr_qf_name || grp_qf_name) { + if (test_opt(sb, USRQUOTA) && usr_qf_name) clear_opt(sb, USRQUOTA); - if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) + if (test_opt(sb, GRPQUOTA) && grp_qf_name) clear_opt(sb, GRPQUOTA); if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { @@ -2030,6 +2044,7 @@ static inline void ext4_show_quota_options(struct seq_file *seq, { #if defined(CONFIG_QUOTA) struct ext4_sb_info *sbi = EXT4_SB(sb); + char *usr_qf_name, *grp_qf_name; if (sbi->s_jquota_fmt) { char *fmtname = ""; @@ -2048,11 +2063,14 @@ static inline void ext4_show_quota_options(struct seq_file *seq, seq_printf(seq, ",jqfmt=%s", fmtname); } - if (sbi->s_qf_names[USRQUOTA]) - seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); - - if (sbi->s_qf_names[GRPQUOTA]) - seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); + rcu_read_lock(); + usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); + grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); + if (usr_qf_name) + seq_show_option(seq, "usrjquota", usr_qf_name); + if (grp_qf_name) + seq_show_option(seq, "grpjquota", grp_qf_name); + rcu_read_unlock(); #endif } @@ -5104,6 +5122,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) int err = 0; #ifdef CONFIG_QUOTA int i, j; + char *to_free[EXT4_MAXQUOTAS]; #endif char *orig_data = kstrdup(data, GFP_KERNEL); @@ -5123,8 +5142,9 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) old_opts.s_jquota_fmt = sbi->s_jquota_fmt; for (i = 0; i < EXT4_MAXQUOTAS; i++) if (sbi->s_qf_names[i]) { - old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], - GFP_KERNEL); + char *qf_name = get_qf_name(sb, sbi, i); + + old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); if (!old_opts.s_qf_names[i]) { for (j = 0; j < i; j++) kfree(old_opts.s_qf_names[j]); @@ -5353,9 +5373,12 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) #ifdef CONFIG_QUOTA sbi->s_jquota_fmt = old_opts.s_jquota_fmt; for (i = 0; i < EXT4_MAXQUOTAS; i++) { - kfree(sbi->s_qf_names[i]); - sbi->s_qf_names[i] = old_opts.s_qf_names[i]; + to_free[i] = get_qf_name(sb, sbi, i); + rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); } + synchronize_rcu(); + for (i = 0; i < EXT4_MAXQUOTAS; i++) + kfree(to_free[i]); #endif kfree(orig_data); return err; @@ -5546,7 +5569,7 @@ static int ext4_write_info(struct super_block *sb, int type) */ static int ext4_quota_on_mount(struct super_block *sb, int type) { - return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], + return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), EXT4_SB(sb)->s_jquota_fmt, type); }

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] ext4: fix use-after-free race in ext4_remount()'s error path" failed to apply to 4.4-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.4-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 33458eaba4dfe778a426df6a19b7aad2ff9f7eec Mon Sep 17 00:00:00 2001 From: Theodore Ts'o <tytso(a)mit.edu> Date: Fri, 12 Oct 2018 09:28:09 -0400 Subject: [PATCH] ext4: fix use-after-free race in ext4_remount()'s error path It's possible for ext4_show_quota_options() to try reading s_qf_names[i] while it is being modified by ext4_remount() --- most notably, in ext4_remount's error path when the original values of the quota file name gets restored. Reported-by: syzbot+a2872d6feea6918008a9(a)syzkaller.appspotmail.com Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Cc: stable(a)kernel.org # 3.2+ diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 86e1bacac757..12f90d48ba61 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -1405,7 +1405,8 @@ struct ext4_sb_info { u32 s_min_batch_time; struct block_device *journal_bdev; #ifdef CONFIG_QUOTA - char *s_qf_names[EXT4_MAXQUOTAS]; /* Names of quota files with journalled quota */ + /* Names of quota files with journalled quota */ + char __rcu *s_qf_names[EXT4_MAXQUOTAS]; int s_jquota_fmt; /* Format of quota to use */ #endif unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */ diff --git a/fs/ext4/super.c b/fs/ext4/super.c index faf293ed8060..a221f1cdf704 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -914,6 +914,18 @@ static inline void ext4_quota_off_umount(struct super_block *sb) for (type = 0; type < EXT4_MAXQUOTAS; type++) ext4_quota_off(sb, type); } + +/* + * This is a helper function which is used in the mount/remount + * codepaths (which holds s_umount) to fetch the quota file name. + */ +static inline char *get_qf_name(struct super_block *sb, + struct ext4_sb_info *sbi, + int type) +{ + return rcu_dereference_protected(sbi->s_qf_names[type], + lockdep_is_held(&sb->s_umount)); +} #else static inline void ext4_quota_off_umount(struct super_block *sb) { @@ -965,7 +977,7 @@ static void ext4_put_super(struct super_block *sb) percpu_free_rwsem(&sbi->s_journal_flag_rwsem); #ifdef CONFIG_QUOTA for (i = 0; i < EXT4_MAXQUOTAS; i++) - kfree(sbi->s_qf_names[i]); + kfree(get_qf_name(sb, sbi, i)); #endif /* Debugging code just in case the in-memory inode orphan list @@ -1531,11 +1543,10 @@ static const char deprecated_msg[] = static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) { struct ext4_sb_info *sbi = EXT4_SB(sb); - char *qname; + char *qname, *old_qname = get_qf_name(sb, sbi, qtype); int ret = -1; - if (sb_any_quota_loaded(sb) && - !sbi->s_qf_names[qtype]) { + if (sb_any_quota_loaded(sb) && !old_qname) { ext4_msg(sb, KERN_ERR, "Cannot change journaled " "quota options when quota turned on"); @@ -1552,8 +1563,8 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) "Not enough memory for storing quotafile name"); return -1; } - if (sbi->s_qf_names[qtype]) { - if (strcmp(sbi->s_qf_names[qtype], qname) == 0) + if (old_qname) { + if (strcmp(old_qname, qname) == 0) ret = 1; else ext4_msg(sb, KERN_ERR, @@ -1566,7 +1577,7 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) "quotafile must be on filesystem root"); goto errout; } - sbi->s_qf_names[qtype] = qname; + rcu_assign_pointer(sbi->s_qf_names[qtype], qname); set_opt(sb, QUOTA); return 1; errout: @@ -1578,15 +1589,16 @@ static int clear_qf_name(struct super_block *sb, int qtype) { struct ext4_sb_info *sbi = EXT4_SB(sb); + char *old_qname = get_qf_name(sb, sbi, qtype); - if (sb_any_quota_loaded(sb) && - sbi->s_qf_names[qtype]) { + if (sb_any_quota_loaded(sb) && old_qname) { ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" " when quota turned on"); return -1; } - kfree(sbi->s_qf_names[qtype]); - sbi->s_qf_names[qtype] = NULL; + rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); + synchronize_rcu(); + kfree(old_qname); return 1; } #endif @@ -1961,7 +1973,7 @@ static int parse_options(char *options, struct super_block *sb, int is_remount) { struct ext4_sb_info *sbi = EXT4_SB(sb); - char *p; + char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; substring_t args[MAX_OPT_ARGS]; int token; @@ -1992,11 +2004,13 @@ static int parse_options(char *options, struct super_block *sb, "Cannot enable project quota enforcement."); return 0; } - if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { - if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) + usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); + grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); + if (usr_qf_name || grp_qf_name) { + if (test_opt(sb, USRQUOTA) && usr_qf_name) clear_opt(sb, USRQUOTA); - if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) + if (test_opt(sb, GRPQUOTA) && grp_qf_name) clear_opt(sb, GRPQUOTA); if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { @@ -2030,6 +2044,7 @@ static inline void ext4_show_quota_options(struct seq_file *seq, { #if defined(CONFIG_QUOTA) struct ext4_sb_info *sbi = EXT4_SB(sb); + char *usr_qf_name, *grp_qf_name; if (sbi->s_jquota_fmt) { char *fmtname = ""; @@ -2048,11 +2063,14 @@ static inline void ext4_show_quota_options(struct seq_file *seq, seq_printf(seq, ",jqfmt=%s", fmtname); } - if (sbi->s_qf_names[USRQUOTA]) - seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); - - if (sbi->s_qf_names[GRPQUOTA]) - seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); + rcu_read_lock(); + usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); + grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); + if (usr_qf_name) + seq_show_option(seq, "usrjquota", usr_qf_name); + if (grp_qf_name) + seq_show_option(seq, "grpjquota", grp_qf_name); + rcu_read_unlock(); #endif } @@ -5104,6 +5122,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) int err = 0; #ifdef CONFIG_QUOTA int i, j; + char *to_free[EXT4_MAXQUOTAS]; #endif char *orig_data = kstrdup(data, GFP_KERNEL); @@ -5123,8 +5142,9 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) old_opts.s_jquota_fmt = sbi->s_jquota_fmt; for (i = 0; i < EXT4_MAXQUOTAS; i++) if (sbi->s_qf_names[i]) { - old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], - GFP_KERNEL); + char *qf_name = get_qf_name(sb, sbi, i); + + old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); if (!old_opts.s_qf_names[i]) { for (j = 0; j < i; j++) kfree(old_opts.s_qf_names[j]); @@ -5353,9 +5373,12 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) #ifdef CONFIG_QUOTA sbi->s_jquota_fmt = old_opts.s_jquota_fmt; for (i = 0; i < EXT4_MAXQUOTAS; i++) { - kfree(sbi->s_qf_names[i]); - sbi->s_qf_names[i] = old_opts.s_qf_names[i]; + to_free[i] = get_qf_name(sb, sbi, i); + rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); } + synchronize_rcu(); + for (i = 0; i < EXT4_MAXQUOTAS; i++) + kfree(to_free[i]); #endif kfree(orig_data); return err; @@ -5546,7 +5569,7 @@ static int ext4_write_info(struct super_block *sb, int type) */ static int ext4_quota_on_mount(struct super_block *sb, int type) { - return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], + return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), EXT4_SB(sb)->s_jquota_fmt, type); }

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] ext4: fix use-after-free race in ext4_remount()'s error path" failed to apply to 4.9-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.9-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 33458eaba4dfe778a426df6a19b7aad2ff9f7eec Mon Sep 17 00:00:00 2001 From: Theodore Ts'o <tytso(a)mit.edu> Date: Fri, 12 Oct 2018 09:28:09 -0400 Subject: [PATCH] ext4: fix use-after-free race in ext4_remount()'s error path It's possible for ext4_show_quota_options() to try reading s_qf_names[i] while it is being modified by ext4_remount() --- most notably, in ext4_remount's error path when the original values of the quota file name gets restored. Reported-by: syzbot+a2872d6feea6918008a9(a)syzkaller.appspotmail.com Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Cc: stable(a)kernel.org # 3.2+ diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 86e1bacac757..12f90d48ba61 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -1405,7 +1405,8 @@ struct ext4_sb_info { u32 s_min_batch_time; struct block_device *journal_bdev; #ifdef CONFIG_QUOTA - char *s_qf_names[EXT4_MAXQUOTAS]; /* Names of quota files with journalled quota */ + /* Names of quota files with journalled quota */ + char __rcu *s_qf_names[EXT4_MAXQUOTAS]; int s_jquota_fmt; /* Format of quota to use */ #endif unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */ diff --git a/fs/ext4/super.c b/fs/ext4/super.c index faf293ed8060..a221f1cdf704 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -914,6 +914,18 @@ static inline void ext4_quota_off_umount(struct super_block *sb) for (type = 0; type < EXT4_MAXQUOTAS; type++) ext4_quota_off(sb, type); } + +/* + * This is a helper function which is used in the mount/remount + * codepaths (which holds s_umount) to fetch the quota file name. + */ +static inline char *get_qf_name(struct super_block *sb, + struct ext4_sb_info *sbi, + int type) +{ + return rcu_dereference_protected(sbi->s_qf_names[type], + lockdep_is_held(&sb->s_umount)); +} #else static inline void ext4_quota_off_umount(struct super_block *sb) { @@ -965,7 +977,7 @@ static void ext4_put_super(struct super_block *sb) percpu_free_rwsem(&sbi->s_journal_flag_rwsem); #ifdef CONFIG_QUOTA for (i = 0; i < EXT4_MAXQUOTAS; i++) - kfree(sbi->s_qf_names[i]); + kfree(get_qf_name(sb, sbi, i)); #endif /* Debugging code just in case the in-memory inode orphan list @@ -1531,11 +1543,10 @@ static const char deprecated_msg[] = static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) { struct ext4_sb_info *sbi = EXT4_SB(sb); - char *qname; + char *qname, *old_qname = get_qf_name(sb, sbi, qtype); int ret = -1; - if (sb_any_quota_loaded(sb) && - !sbi->s_qf_names[qtype]) { + if (sb_any_quota_loaded(sb) && !old_qname) { ext4_msg(sb, KERN_ERR, "Cannot change journaled " "quota options when quota turned on"); @@ -1552,8 +1563,8 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) "Not enough memory for storing quotafile name"); return -1; } - if (sbi->s_qf_names[qtype]) { - if (strcmp(sbi->s_qf_names[qtype], qname) == 0) + if (old_qname) { + if (strcmp(old_qname, qname) == 0) ret = 1; else ext4_msg(sb, KERN_ERR, @@ -1566,7 +1577,7 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) "quotafile must be on filesystem root"); goto errout; } - sbi->s_qf_names[qtype] = qname; + rcu_assign_pointer(sbi->s_qf_names[qtype], qname); set_opt(sb, QUOTA); return 1; errout: @@ -1578,15 +1589,16 @@ static int clear_qf_name(struct super_block *sb, int qtype) { struct ext4_sb_info *sbi = EXT4_SB(sb); + char *old_qname = get_qf_name(sb, sbi, qtype); - if (sb_any_quota_loaded(sb) && - sbi->s_qf_names[qtype]) { + if (sb_any_quota_loaded(sb) && old_qname) { ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" " when quota turned on"); return -1; } - kfree(sbi->s_qf_names[qtype]); - sbi->s_qf_names[qtype] = NULL; + rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); + synchronize_rcu(); + kfree(old_qname); return 1; } #endif @@ -1961,7 +1973,7 @@ static int parse_options(char *options, struct super_block *sb, int is_remount) { struct ext4_sb_info *sbi = EXT4_SB(sb); - char *p; + char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; substring_t args[MAX_OPT_ARGS]; int token; @@ -1992,11 +2004,13 @@ static int parse_options(char *options, struct super_block *sb, "Cannot enable project quota enforcement."); return 0; } - if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { - if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) + usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); + grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); + if (usr_qf_name || grp_qf_name) { + if (test_opt(sb, USRQUOTA) && usr_qf_name) clear_opt(sb, USRQUOTA); - if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) + if (test_opt(sb, GRPQUOTA) && grp_qf_name) clear_opt(sb, GRPQUOTA); if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { @@ -2030,6 +2044,7 @@ static inline void ext4_show_quota_options(struct seq_file *seq, { #if defined(CONFIG_QUOTA) struct ext4_sb_info *sbi = EXT4_SB(sb); + char *usr_qf_name, *grp_qf_name; if (sbi->s_jquota_fmt) { char *fmtname = ""; @@ -2048,11 +2063,14 @@ static inline void ext4_show_quota_options(struct seq_file *seq, seq_printf(seq, ",jqfmt=%s", fmtname); } - if (sbi->s_qf_names[USRQUOTA]) - seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); - - if (sbi->s_qf_names[GRPQUOTA]) - seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); + rcu_read_lock(); + usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); + grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); + if (usr_qf_name) + seq_show_option(seq, "usrjquota", usr_qf_name); + if (grp_qf_name) + seq_show_option(seq, "grpjquota", grp_qf_name); + rcu_read_unlock(); #endif } @@ -5104,6 +5122,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) int err = 0; #ifdef CONFIG_QUOTA int i, j; + char *to_free[EXT4_MAXQUOTAS]; #endif char *orig_data = kstrdup(data, GFP_KERNEL); @@ -5123,8 +5142,9 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) old_opts.s_jquota_fmt = sbi->s_jquota_fmt; for (i = 0; i < EXT4_MAXQUOTAS; i++) if (sbi->s_qf_names[i]) { - old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], - GFP_KERNEL); + char *qf_name = get_qf_name(sb, sbi, i); + + old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); if (!old_opts.s_qf_names[i]) { for (j = 0; j < i; j++) kfree(old_opts.s_qf_names[j]); @@ -5353,9 +5373,12 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) #ifdef CONFIG_QUOTA sbi->s_jquota_fmt = old_opts.s_jquota_fmt; for (i = 0; i < EXT4_MAXQUOTAS; i++) { - kfree(sbi->s_qf_names[i]); - sbi->s_qf_names[i] = old_opts.s_qf_names[i]; + to_free[i] = get_qf_name(sb, sbi, i); + rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); } + synchronize_rcu(); + for (i = 0; i < EXT4_MAXQUOTAS; i++) + kfree(to_free[i]); #endif kfree(orig_data); return err; @@ -5546,7 +5569,7 @@ static int ext4_write_info(struct super_block *sb, int type) */ static int ext4_quota_on_mount(struct super_block *sb, int type) { - return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], + return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), EXT4_SB(sb)->s_jquota_fmt, type); }

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] IB/rxe: Revise the ib_wr_opcode enum" failed to apply to 4.9-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.9-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 9a59739bd01f77db6fbe2955a4fce165f0f43568 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe <jgg(a)mellanox.com> Date: Tue, 14 Aug 2018 15:33:02 -0700 Subject: [PATCH] IB/rxe: Revise the ib_wr_opcode enum This enum has become part of the uABI, as both RXE and the ib_uverbs_post_send() command expect userspace to supply values from this enum. So it should be properly placed in include/uapi/rdma. In userspace this enum is called 'enum ibv_wr_opcode' as part of libibverbs.h. That enum defines different values for IB_WR_LOCAL_INV, IB_WR_SEND_WITH_INV, and IB_WR_LSO. These were introduced (incorrectly, it turns out) into libiberbs in 2015. The kernel has changed its mind on the numbering for several of the IB_WC values over the years, but has remained stable on IB_WR_LOCAL_INV and below. Based on this we can conclude that there is no real user space user of the values beyond IB_WR_ATOMIC_FETCH_AND_ADD, as they have never worked via rdma-core. This is confirmed by inspection, only rxe uses the kernel enum and implements the latter operations. rxe has clearly never worked with these attributes from userspace. Other drivers that support these opcodes implement the functionality without calling out to the kernel. To make IB_WR_SEND_WITH_INV and related work for RXE in userspace we choose to renumber the IB_WR enum in the kernel to match the uABI that userspace has bee using since before Soft RoCE was merged. This is an overall simpler configuration for the whole software stack, and obviously can't break anything existing. Reported-by: Seth Howell <seth.howell(a)intel.com> Tested-by: Seth Howell <seth.howell(a)intel.com> Fixes: 8700e3e7c485 ("Soft RoCE driver") Cc: <stable(a)vger.kernel.org> Signed-off-by: Jason Gunthorpe <jgg(a)mellanox.com> diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 6076c9b72ab9..e463d3007a35 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1281,21 +1281,27 @@ struct ib_qp_attr { }; enum ib_wr_opcode { - IB_WR_RDMA_WRITE, - IB_WR_RDMA_WRITE_WITH_IMM, - IB_WR_SEND, - IB_WR_SEND_WITH_IMM, - IB_WR_RDMA_READ, - IB_WR_ATOMIC_CMP_AND_SWP, - IB_WR_ATOMIC_FETCH_AND_ADD, - IB_WR_LSO, - IB_WR_SEND_WITH_INV, - IB_WR_RDMA_READ_WITH_INV, - IB_WR_LOCAL_INV, - IB_WR_REG_MR, - IB_WR_MASKED_ATOMIC_CMP_AND_SWP, - IB_WR_MASKED_ATOMIC_FETCH_AND_ADD, + /* These are shared with userspace */ + IB_WR_RDMA_WRITE = IB_UVERBS_WR_RDMA_WRITE, + IB_WR_RDMA_WRITE_WITH_IMM = IB_UVERBS_WR_RDMA_WRITE_WITH_IMM, + IB_WR_SEND = IB_UVERBS_WR_SEND, + IB_WR_SEND_WITH_IMM = IB_UVERBS_WR_SEND_WITH_IMM, + IB_WR_RDMA_READ = IB_UVERBS_WR_RDMA_READ, + IB_WR_ATOMIC_CMP_AND_SWP = IB_UVERBS_WR_ATOMIC_CMP_AND_SWP, + IB_WR_ATOMIC_FETCH_AND_ADD = IB_UVERBS_WR_ATOMIC_FETCH_AND_ADD, + IB_WR_LSO = IB_UVERBS_WR_TSO, + IB_WR_SEND_WITH_INV = IB_UVERBS_WR_SEND_WITH_INV, + IB_WR_RDMA_READ_WITH_INV = IB_UVERBS_WR_RDMA_READ_WITH_INV, + IB_WR_LOCAL_INV = IB_UVERBS_WR_LOCAL_INV, + IB_WR_MASKED_ATOMIC_CMP_AND_SWP = + IB_UVERBS_WR_MASKED_ATOMIC_CMP_AND_SWP, + IB_WR_MASKED_ATOMIC_FETCH_AND_ADD = + IB_UVERBS_WR_MASKED_ATOMIC_FETCH_AND_ADD, + + /* These are kernel only and can not be issued by userspace */ + IB_WR_REG_MR = 0x20, IB_WR_REG_SIG_MR, + /* reserve values for low level drivers' internal use. * These values will not be used at all in the ib core layer. */ diff --git a/include/uapi/rdma/ib_user_verbs.h b/include/uapi/rdma/ib_user_verbs.h index 25a16760de2a..1254b51a551a 100644 --- a/include/uapi/rdma/ib_user_verbs.h +++ b/include/uapi/rdma/ib_user_verbs.h @@ -763,10 +763,28 @@ struct ib_uverbs_sge { __u32 lkey; }; +enum ib_uverbs_wr_opcode { + IB_UVERBS_WR_RDMA_WRITE = 0, + IB_UVERBS_WR_RDMA_WRITE_WITH_IMM = 1, + IB_UVERBS_WR_SEND = 2, + IB_UVERBS_WR_SEND_WITH_IMM = 3, + IB_UVERBS_WR_RDMA_READ = 4, + IB_UVERBS_WR_ATOMIC_CMP_AND_SWP = 5, + IB_UVERBS_WR_ATOMIC_FETCH_AND_ADD = 6, + IB_UVERBS_WR_LOCAL_INV = 7, + IB_UVERBS_WR_BIND_MW = 8, + IB_UVERBS_WR_SEND_WITH_INV = 9, + IB_UVERBS_WR_TSO = 10, + IB_UVERBS_WR_RDMA_READ_WITH_INV = 11, + IB_UVERBS_WR_MASKED_ATOMIC_CMP_AND_SWP = 12, + IB_UVERBS_WR_MASKED_ATOMIC_FETCH_AND_ADD = 13, + /* Review enum ib_wr_opcode before modifying this */ +}; + struct ib_uverbs_send_wr { __aligned_u64 wr_id; __u32 num_sge; - __u32 opcode; + __u32 opcode; /* see enum ib_uverbs_wr_opcode */ __u32 send_flags; union { __be32 imm_data;

6 years, 2 months

1
0
0 0

FAILED: patch "[PATCH] IB/rxe: Revise the ib_wr_opcode enum" failed to apply to 4.14-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.14-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 9a59739bd01f77db6fbe2955a4fce165f0f43568 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe <jgg(a)mellanox.com> Date: Tue, 14 Aug 2018 15:33:02 -0700 Subject: [PATCH] IB/rxe: Revise the ib_wr_opcode enum This enum has become part of the uABI, as both RXE and the ib_uverbs_post_send() command expect userspace to supply values from this enum. So it should be properly placed in include/uapi/rdma. In userspace this enum is called 'enum ibv_wr_opcode' as part of libibverbs.h. That enum defines different values for IB_WR_LOCAL_INV, IB_WR_SEND_WITH_INV, and IB_WR_LSO. These were introduced (incorrectly, it turns out) into libiberbs in 2015. The kernel has changed its mind on the numbering for several of the IB_WC values over the years, but has remained stable on IB_WR_LOCAL_INV and below. Based on this we can conclude that there is no real user space user of the values beyond IB_WR_ATOMIC_FETCH_AND_ADD, as they have never worked via rdma-core. This is confirmed by inspection, only rxe uses the kernel enum and implements the latter operations. rxe has clearly never worked with these attributes from userspace. Other drivers that support these opcodes implement the functionality without calling out to the kernel. To make IB_WR_SEND_WITH_INV and related work for RXE in userspace we choose to renumber the IB_WR enum in the kernel to match the uABI that userspace has bee using since before Soft RoCE was merged. This is an overall simpler configuration for the whole software stack, and obviously can't break anything existing. Reported-by: Seth Howell <seth.howell(a)intel.com> Tested-by: Seth Howell <seth.howell(a)intel.com> Fixes: 8700e3e7c485 ("Soft RoCE driver") Cc: <stable(a)vger.kernel.org> Signed-off-by: Jason Gunthorpe <jgg(a)mellanox.com> diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 6076c9b72ab9..e463d3007a35 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1281,21 +1281,27 @@ struct ib_qp_attr { }; enum ib_wr_opcode { - IB_WR_RDMA_WRITE, - IB_WR_RDMA_WRITE_WITH_IMM, - IB_WR_SEND, - IB_WR_SEND_WITH_IMM, - IB_WR_RDMA_READ, - IB_WR_ATOMIC_CMP_AND_SWP, - IB_WR_ATOMIC_FETCH_AND_ADD, - IB_WR_LSO, - IB_WR_SEND_WITH_INV, - IB_WR_RDMA_READ_WITH_INV, - IB_WR_LOCAL_INV, - IB_WR_REG_MR, - IB_WR_MASKED_ATOMIC_CMP_AND_SWP, - IB_WR_MASKED_ATOMIC_FETCH_AND_ADD, + /* These are shared with userspace */ + IB_WR_RDMA_WRITE = IB_UVERBS_WR_RDMA_WRITE, + IB_WR_RDMA_WRITE_WITH_IMM = IB_UVERBS_WR_RDMA_WRITE_WITH_IMM, + IB_WR_SEND = IB_UVERBS_WR_SEND, + IB_WR_SEND_WITH_IMM = IB_UVERBS_WR_SEND_WITH_IMM, + IB_WR_RDMA_READ = IB_UVERBS_WR_RDMA_READ, + IB_WR_ATOMIC_CMP_AND_SWP = IB_UVERBS_WR_ATOMIC_CMP_AND_SWP, + IB_WR_ATOMIC_FETCH_AND_ADD = IB_UVERBS_WR_ATOMIC_FETCH_AND_ADD, + IB_WR_LSO = IB_UVERBS_WR_TSO, + IB_WR_SEND_WITH_INV = IB_UVERBS_WR_SEND_WITH_INV, + IB_WR_RDMA_READ_WITH_INV = IB_UVERBS_WR_RDMA_READ_WITH_INV, + IB_WR_LOCAL_INV = IB_UVERBS_WR_LOCAL_INV, + IB_WR_MASKED_ATOMIC_CMP_AND_SWP = + IB_UVERBS_WR_MASKED_ATOMIC_CMP_AND_SWP, + IB_WR_MASKED_ATOMIC_FETCH_AND_ADD = + IB_UVERBS_WR_MASKED_ATOMIC_FETCH_AND_ADD, + + /* These are kernel only and can not be issued by userspace */ + IB_WR_REG_MR = 0x20, IB_WR_REG_SIG_MR, + /* reserve values for low level drivers' internal use. * These values will not be used at all in the ib core layer. */ diff --git a/include/uapi/rdma/ib_user_verbs.h b/include/uapi/rdma/ib_user_verbs.h index 25a16760de2a..1254b51a551a 100644 --- a/include/uapi/rdma/ib_user_verbs.h +++ b/include/uapi/rdma/ib_user_verbs.h @@ -763,10 +763,28 @@ struct ib_uverbs_sge { __u32 lkey; }; +enum ib_uverbs_wr_opcode { + IB_UVERBS_WR_RDMA_WRITE = 0, + IB_UVERBS_WR_RDMA_WRITE_WITH_IMM = 1, + IB_UVERBS_WR_SEND = 2, + IB_UVERBS_WR_SEND_WITH_IMM = 3, + IB_UVERBS_WR_RDMA_READ = 4, + IB_UVERBS_WR_ATOMIC_CMP_AND_SWP = 5, + IB_UVERBS_WR_ATOMIC_FETCH_AND_ADD = 6, + IB_UVERBS_WR_LOCAL_INV = 7, + IB_UVERBS_WR_BIND_MW = 8, + IB_UVERBS_WR_SEND_WITH_INV = 9, + IB_UVERBS_WR_TSO = 10, + IB_UVERBS_WR_RDMA_READ_WITH_INV = 11, + IB_UVERBS_WR_MASKED_ATOMIC_CMP_AND_SWP = 12, + IB_UVERBS_WR_MASKED_ATOMIC_FETCH_AND_ADD = 13, + /* Review enum ib_wr_opcode before modifying this */ +}; + struct ib_uverbs_send_wr { __aligned_u64 wr_id; __u32 num_sge; - __u32 opcode; + __u32 opcode; /* see enum ib_uverbs_wr_opcode */ __u32 send_flags; union { __be32 imm_data;

6 years, 2 months

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror November 2018