From: Linus Walleij linus.walleij@linaro.org
[ Upstream commit d6e81532b10d8deb2bc30f7b44f09534876893e3 ]
Making virt_to_pfn() a static inline taking a strongly typed (const void *) makes the contract of a passing a pointer of that type to the function explicit and exposes any misuse of the macro virt_to_pfn() acting polymorphic and accepting many types such as (void *), (unitptr_t) or (unsigned long) as arguments without warnings.
For symmetry do the same with pfn_to_virt().
For compiletime resolution of __pa() we need PAGE_OFFSET which was not available to __pa() and resolved by the preprocessor wherever __pa() was used. Fix this by explicitly including <asm/mem-layout.h> where required, following the pattern of the architectures page.h file.
Acked-by: Brian Cain bcain@quicinc.com Signed-off-by: Linus Walleij linus.walleij@linaro.org Signed-off-by: Arnd Bergmann arnd@arndb.de Signed-off-by: Sasha Levin sashal@kernel.org --- arch/hexagon/include/asm/page.h | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)
diff --git a/arch/hexagon/include/asm/page.h b/arch/hexagon/include/asm/page.h index 7cbf719c578e..2d8c681c3469 100644 --- a/arch/hexagon/include/asm/page.h +++ b/arch/hexagon/include/asm/page.h @@ -78,6 +78,9 @@ typedef struct page *pgtable_t; #define __pgd(x) ((pgd_t) { (x) }) #define __pgprot(x) ((pgprot_t) { (x) })
+/* Needed for PAGE_OFFSET used in the macro right below */ +#include <asm/mem-layout.h> + /* * We need a __pa and a __va routine for kernel space. * MIPS says they're only used during mem_init. @@ -126,8 +129,16 @@ static inline void clear_page(void *page) */ #define page_to_phys(page) (page_to_pfn(page) << PAGE_SHIFT)
-#define virt_to_pfn(kaddr) (__pa(kaddr) >> PAGE_SHIFT) -#define pfn_to_virt(pfn) __va((pfn) << PAGE_SHIFT) +static inline unsigned long virt_to_pfn(const void *kaddr) +{ + return __pa(kaddr) >> PAGE_SHIFT; +} + +static inline void *pfn_to_virt(unsigned long pfn) +{ + return (void *)((unsigned long)__va(pfn) << PAGE_SHIFT); +} +
#define page_to_virt(page) __va(page_to_phys(page))
From: Heiko Carstens hca@linux.ibm.com
[ Upstream commit 8b13601d19c541158a6e18b278c00ba69ae37829 ]
If the content of the floating point control (fpc) register of a traced process is modified with the ptrace interface the new value is tested for validity by temporarily loading it into the fpc register.
This may lead to corruption of the fpc register of the tracing process: if an interrupt happens while the value is temporarily loaded into the fpc register, and within interrupt context floating point or vector registers are used, the current fp/vx registers are saved with save_fpu_regs() assuming they belong to user space and will be loaded into fp/vx registers when returning to user space.
test_fp_ctl() restores the original user space fpc register value, however it will be discarded, when returning to user space.
In result the tracer will incorrectly continue to run with the value that was supposed to be used for the traced process.
Fix this by saving fpu register contents with save_fpu_regs() before using test_fp_ctl().
Reviewed-by: Claudio Imbrenda imbrenda@linux.ibm.com Signed-off-by: Heiko Carstens hca@linux.ibm.com Signed-off-by: Alexander Gordeev agordeev@linux.ibm.com Signed-off-by: Sasha Levin sashal@kernel.org --- arch/s390/kernel/ptrace.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c index 3009bb527252..f381caddd905 100644 --- a/arch/s390/kernel/ptrace.c +++ b/arch/s390/kernel/ptrace.c @@ -411,6 +411,7 @@ static int __poke_user(struct task_struct *child, addr_t addr, addr_t data) /* * floating point control reg. is in the thread structure */ + save_fpu_regs(); if ((unsigned int) data != 0 || test_fp_ctl(data >> (BITS_PER_LONG - 32))) return -EINVAL; @@ -771,6 +772,7 @@ static int __poke_user_compat(struct task_struct *child, /* * floating point control reg. is in the thread structure */ + save_fpu_regs(); if (test_fp_ctl(tmp)) return -EINVAL; child->thread.fpu.fpc = data; @@ -1010,9 +1012,7 @@ static int s390_fpregs_set(struct task_struct *target, int rc = 0; freg_t fprs[__NUM_FPRS];
- if (target == current) - save_fpu_regs(); - + save_fpu_regs(); if (MACHINE_HAS_VX) convert_vx_to_fp(fprs, target->thread.fpu.vxrs); else
From: Heiko Carstens hca@linux.ibm.com
[ Upstream commit b988b1bb0053c0dcd26187d29ef07566a565cf55 ]
kvm_arch_vcpu_ioctl_set_fpu() allows to set the floating point control (fpc) register of a guest cpu. The new value is tested for validity by temporarily loading it into the fpc register.
This may lead to corruption of the fpc register of the host process: if an interrupt happens while the value is temporarily loaded into the fpc register, and within interrupt context floating point or vector registers are used, the current fp/vx registers are saved with save_fpu_regs() assuming they belong to user space and will be loaded into fp/vx registers when returning to user space.
test_fp_ctl() restores the original user space / host process fpc register value, however it will be discarded, when returning to user space.
In result the host process will incorrectly continue to run with the value that was supposed to be used for a guest cpu.
Fix this by simply removing the test. There is another test right before the SIE context is entered which will handles invalid values.
This results in a change of behaviour: invalid values will now be accepted instead of that the ioctl fails with -EINVAL. This seems to be acceptable, given that this interface is most likely not used anymore, and this is in addition the same behaviour implemented with the memory mapped interface (replace invalid values with zero) - see sync_regs() in kvm-s390.c.
Reviewed-by: Christian Borntraeger borntraeger@linux.ibm.com Reviewed-by: Claudio Imbrenda imbrenda@linux.ibm.com Signed-off-by: Heiko Carstens hca@linux.ibm.com Signed-off-by: Alexander Gordeev agordeev@linux.ibm.com Signed-off-by: Sasha Levin sashal@kernel.org --- arch/s390/kvm/kvm-s390.c | 5 ----- 1 file changed, 5 deletions(-)
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c index 7a326d03087a..f6c27b44766f 100644 --- a/arch/s390/kvm/kvm-s390.c +++ b/arch/s390/kvm/kvm-s390.c @@ -3649,10 +3649,6 @@ int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
vcpu_load(vcpu);
- if (test_fp_ctl(fpu->fpc)) { - ret = -EINVAL; - goto out; - } vcpu->run->s.regs.fpc = fpu->fpc; if (MACHINE_HAS_VX) convert_fp_to_vx((__vector128 *) vcpu->run->s.regs.vrs, @@ -3660,7 +3656,6 @@ int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) else memcpy(vcpu->run->s.regs.fprs, &fpu->fprs, sizeof(fpu->fprs));
-out: vcpu_put(vcpu); return ret; }
From: Anna Schumaker Anna.Schumaker@Netapp.com
[ Upstream commit 31b62908693c90d4d07db597e685d9f25a120073 ]
I received the following warning while running cthon against an ontap server running pNFS:
[ 57.202521] ============================= [ 57.202522] WARNING: suspicious RCU usage [ 57.202523] 6.7.0-rc3-g2cc14f52aeb7 #41492 Not tainted [ 57.202525] ----------------------------- [ 57.202525] net/sunrpc/xprtmultipath.c:349 RCU-list traversed in non-reader section!! [ 57.202527] other info that might help us debug this:
[ 57.202528] rcu_scheduler_active = 2, debug_locks = 1 [ 57.202529] no locks held by test5/3567. [ 57.202530] stack backtrace: [ 57.202532] CPU: 0 PID: 3567 Comm: test5 Not tainted 6.7.0-rc3-g2cc14f52aeb7 #41492 5b09971b4965c0aceba19f3eea324a4a806e227e [ 57.202534] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS unknown 2/2/2022 [ 57.202536] Call Trace: [ 57.202537] <TASK> [ 57.202540] dump_stack_lvl+0x77/0xb0 [ 57.202551] lockdep_rcu_suspicious+0x154/0x1a0 [ 57.202556] rpc_xprt_switch_has_addr+0x17c/0x190 [sunrpc ebe02571b9a8ceebf7d98e71675af20c19bdb1f6] [ 57.202596] rpc_clnt_setup_test_and_add_xprt+0x50/0x180 [sunrpc ebe02571b9a8ceebf7d98e71675af20c19bdb1f6] [ 57.202621] ? rpc_clnt_add_xprt+0x254/0x300 [sunrpc ebe02571b9a8ceebf7d98e71675af20c19bdb1f6] [ 57.202646] rpc_clnt_add_xprt+0x27a/0x300 [sunrpc ebe02571b9a8ceebf7d98e71675af20c19bdb1f6] [ 57.202671] ? __pfx_rpc_clnt_setup_test_and_add_xprt+0x10/0x10 [sunrpc ebe02571b9a8ceebf7d98e71675af20c19bdb1f6] [ 57.202696] nfs4_pnfs_ds_connect+0x345/0x760 [nfsv4 c716d88496ded0ea6d289bbea684fa996f9b57a9] [ 57.202728] ? __pfx_nfs4_test_session_trunk+0x10/0x10 [nfsv4 c716d88496ded0ea6d289bbea684fa996f9b57a9] [ 57.202754] nfs4_fl_prepare_ds+0x75/0xc0 [nfs_layout_nfsv41_files e3a4187f18ae8a27b630f9feae6831b584a9360a] [ 57.202760] filelayout_write_pagelist+0x4a/0x200 [nfs_layout_nfsv41_files e3a4187f18ae8a27b630f9feae6831b584a9360a] [ 57.202765] pnfs_generic_pg_writepages+0xbe/0x230 [nfsv4 c716d88496ded0ea6d289bbea684fa996f9b57a9] [ 57.202788] __nfs_pageio_add_request+0x3fd/0x520 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202813] nfs_pageio_add_request+0x18b/0x390 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202831] nfs_do_writepage+0x116/0x1e0 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202849] nfs_writepages_callback+0x13/0x30 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202866] write_cache_pages+0x265/0x450 [ 57.202870] ? __pfx_nfs_writepages_callback+0x10/0x10 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202891] nfs_writepages+0x141/0x230 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202913] do_writepages+0xd2/0x230 [ 57.202917] ? filemap_fdatawrite_wbc+0x5c/0x80 [ 57.202921] filemap_fdatawrite_wbc+0x67/0x80 [ 57.202924] filemap_write_and_wait_range+0xd9/0x170 [ 57.202930] nfs_wb_all+0x49/0x180 [nfs 6c976fa593a7c2976f5a0aeb4965514a828e6902] [ 57.202947] nfs4_file_flush+0x72/0xb0 [nfsv4 c716d88496ded0ea6d289bbea684fa996f9b57a9] [ 57.202969] __se_sys_close+0x46/0xd0 [ 57.202972] do_syscall_64+0x68/0x100 [ 57.202975] ? do_syscall_64+0x77/0x100 [ 57.202976] ? do_syscall_64+0x77/0x100 [ 57.202979] entry_SYSCALL_64_after_hwframe+0x6e/0x76 [ 57.202982] RIP: 0033:0x7fe2b12e4a94 [ 57.202985] Code: 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 80 3d d5 18 0e 00 00 74 13 b8 03 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 44 c3 0f 1f 00 48 83 ec 18 89 7c 24 0c e8 c3 [ 57.202987] RSP: 002b:00007ffe857ddb38 EFLAGS: 00000202 ORIG_RAX: 0000000000000003 [ 57.202989] RAX: ffffffffffffffda RBX: 00007ffe857dfd68 RCX: 00007fe2b12e4a94 [ 57.202991] RDX: 0000000000002000 RSI: 00007ffe857ddc40 RDI: 0000000000000003 [ 57.202992] RBP: 00007ffe857dfc50 R08: 7fffffffffffffff R09: 0000000065650f49 [ 57.202993] R10: 00007fe2b11f8300 R11: 0000000000000202 R12: 0000000000000000 [ 57.202994] R13: 00007ffe857dfd80 R14: 00007fe2b1445000 R15: 0000000000000000 [ 57.202999] </TASK>
The problem seems to be that two out of three callers aren't taking the rcu_read_lock() before calling the list_for_each_entry_rcu() function in rpc_xprt_switch_has_addr(). I fix this by having rpc_xprt_switch_has_addr() unconditionaly take the rcu_read_lock(), which is okay to do recursively in the case that the lock has already been taken by a caller.
Reviewed-by: Jeff Layton jlayton@kernel.org Signed-off-by: Anna Schumaker Anna.Schumaker@Netapp.com Signed-off-by: Sasha Levin sashal@kernel.org --- net/sunrpc/xprtmultipath.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/net/sunrpc/xprtmultipath.c b/net/sunrpc/xprtmultipath.c index 78c075a68c04..a11e80d17830 100644 --- a/net/sunrpc/xprtmultipath.c +++ b/net/sunrpc/xprtmultipath.c @@ -253,8 +253,9 @@ struct rpc_xprt *xprt_iter_current_entry(struct rpc_xprt_iter *xpi) return xprt_switch_find_current_entry(head, xpi->xpi_cursor); }
-bool rpc_xprt_switch_has_addr(struct rpc_xprt_switch *xps, - const struct sockaddr *sap) +static +bool __rpc_xprt_switch_has_addr(struct rpc_xprt_switch *xps, + const struct sockaddr *sap) { struct list_head *head; struct rpc_xprt *pos; @@ -273,6 +274,18 @@ bool rpc_xprt_switch_has_addr(struct rpc_xprt_switch *xps, return false; }
+bool rpc_xprt_switch_has_addr(struct rpc_xprt_switch *xps, + const struct sockaddr *sap) +{ + bool res; + + rcu_read_lock(); + res = __rpc_xprt_switch_has_addr(xps, sap); + rcu_read_unlock(); + + return res; +} + static struct rpc_xprt *xprt_switch_find_next_entry(struct list_head *head, const struct rpc_xprt *cur)
From: Gabriel Krisman Bertazi krisman@suse.de
[ Upstream commit cd72c7ef5fed44272272a105b1da22810c91be69 ]
Even though it seems to be able to resolve some names of case-insensitive directories, the lack of d_hash and d_compare means we end up with a broken state in the d_cache. Considering it was never a goal to support these two together, and we are preparing to use d_revalidate in case-insensitive filesystems, which would make the combination even more broken, reject any attempt to get a casefolded inode from ecryptfs.
Signed-off-by: Gabriel Krisman Bertazi krisman@suse.de Reviewed-by: Eric Biggers ebiggers@google.com Signed-off-by: Sasha Levin sashal@kernel.org --- fs/ecryptfs/inode.c | 8 ++++++++ 1 file changed, 8 insertions(+)
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c index e23752d9a79f..c867a0d62f36 100644 --- a/fs/ecryptfs/inode.c +++ b/fs/ecryptfs/inode.c @@ -76,6 +76,14 @@ static struct inode *__ecryptfs_get_inode(struct inode *lower_inode,
if (lower_inode->i_sb != ecryptfs_superblock_to_lower(sb)) return ERR_PTR(-EXDEV); + + /* Reject dealing with casefold directories. */ + if (IS_CASEFOLDED(lower_inode)) { + pr_err_ratelimited("%s: Can't handle casefolded directory.\n", + __func__); + return ERR_PTR(-EREMOTE); + } + if (!igrab(lower_inode)) return ERR_PTR(-ESTALE); inode = iget5_locked(sb, (unsigned long)lower_inode,
From: Ye Bin yebin10@huawei.com
[ Upstream commit 68da4c44b994aea797eb9821acb3a4a36015293e ]
Suppose we issue two FITRIM ioctls for ranges [0,15] and [16,31] with mininum length of trimmed range set to 8 blocks. If we have say a range of blocks 10-22 free, this range will not be trimmed because it straddles the boundary of the two FITRIM ranges and neither part is big enough. This is a bit surprising to some users that call FITRIM on smaller ranges of blocks to limit impact on the system. Also XFS trims all free space extents that overlap with the specified range so we are inconsistent among filesystems. Let's change ext4_try_to_trim_range() to consider for trimming the whole free space extent that straddles the end of specified range, not just the part of it within the range.
Signed-off-by: Ye Bin yebin10@huawei.com Reviewed-by: Jan Kara jack@suse.cz Link: https://lore.kernel.org/r/20231216010919.1995851-1-yebin10@huawei.com Signed-off-by: Theodore Ts'o tytso@mit.edu Signed-off-by: Sasha Levin sashal@kernel.org --- fs/ext4/mballoc.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index 3babc07ae613..b9252fb2253e 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -5911,13 +5911,15 @@ static int ext4_try_to_trim_range(struct super_block *sb, struct ext4_buddy *e4b, ext4_grpblk_t start, ext4_grpblk_t max, ext4_grpblk_t minblocks) { - ext4_grpblk_t next, count, free_count; + ext4_grpblk_t next, count, free_count, last, origin_start; bool set_trimmed = false; void *bitmap;
+ last = ext4_last_grp_cluster(sb, e4b->bd_group); bitmap = e4b->bd_bitmap; - if (start == 0 && max >= ext4_last_grp_cluster(sb, e4b->bd_group)) + if (start == 0 && max >= last) set_trimmed = true; + origin_start = start; start = max(e4b->bd_info->bb_first_free, start); count = 0; free_count = 0; @@ -5926,7 +5928,10 @@ static int ext4_try_to_trim_range(struct super_block *sb, start = mb_find_next_zero_bit(bitmap, max + 1, start); if (start > max) break; - next = mb_find_next_bit(bitmap, max + 1, start); + + next = mb_find_next_bit(bitmap, last + 1, start); + if (origin_start == 0 && next >= last) + set_trimmed = true;
if ((next - start) >= minblocks) { int ret = ext4_trim_extent(sb, start, next - start, e4b);
From: Baokun Li libaokun1@huawei.com
[ Upstream commit 658a52344fb139f9531e7543a6e0015b630feb38 ]
The maximum value of flexbg_size is 2^31, but the maximum value of int is (2^31 - 1), so overflow may occur when the type of flexbg_size is declared as int.
For example, when uninit_mask is initialized in ext4_alloc_group_tables(), if flexbg_size == 2^31, the initialized uninit_mask is incorrect, and this may causes set_flexbg_block_bitmap() to trigger a BUG_ON().
Therefore, the flexbg_size type is declared as unsigned int to avoid overflow and memory waste.
Signed-off-by: Baokun Li libaokun1@huawei.com Reviewed-by: Jan Kara jack@suse.cz Link: https://lore.kernel.org/r/20231023013057.2117948-2-libaokun1@huawei.com Signed-off-by: Theodore Ts'o tytso@mit.edu Signed-off-by: Sasha Levin sashal@kernel.org --- fs/ext4/resize.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-)
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c index 9b4199a1e039..96d278688fd7 100644 --- a/fs/ext4/resize.c +++ b/fs/ext4/resize.c @@ -237,7 +237,7 @@ struct ext4_new_flex_group_data { * * Returns NULL on failure otherwise address of the allocated structure. */ -static struct ext4_new_flex_group_data *alloc_flex_gd(unsigned long flexbg_size) +static struct ext4_new_flex_group_data *alloc_flex_gd(unsigned int flexbg_size) { struct ext4_new_flex_group_data *flex_gd;
@@ -292,7 +292,7 @@ static void free_flex_gd(struct ext4_new_flex_group_data *flex_gd) */ static int ext4_alloc_group_tables(struct super_block *sb, struct ext4_new_flex_group_data *flex_gd, - int flexbg_size) + unsigned int flexbg_size) { struct ext4_new_group_data *group_data = flex_gd->groups; ext4_fsblk_t start_blk; @@ -393,12 +393,12 @@ static int ext4_alloc_group_tables(struct super_block *sb, group = group_data[0].group;
printk(KERN_DEBUG "EXT4-fs: adding a flex group with " - "%d groups, flexbg size is %d:\n", flex_gd->count, + "%u groups, flexbg size is %u:\n", flex_gd->count, flexbg_size);
for (i = 0; i < flex_gd->count; i++) { ext4_debug( - "adding %s group %u: %u blocks (%d free, %d mdata blocks)\n", + "adding %s group %u: %u blocks (%u free, %u mdata blocks)\n", ext4_bg_has_super(sb, group + i) ? "normal" : "no-super", group + i, group_data[i].blocks_count, @@ -1563,7 +1563,7 @@ static int ext4_flex_group_add(struct super_block *sb, static int ext4_setup_next_flex_gd(struct super_block *sb, struct ext4_new_flex_group_data *flex_gd, ext4_fsblk_t n_blocks_count, - unsigned long flexbg_size) + unsigned int flexbg_size) { struct ext4_sb_info *sbi = EXT4_SB(sb); struct ext4_super_block *es = sbi->s_es; @@ -1941,8 +1941,9 @@ int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count) ext4_fsblk_t o_blocks_count; ext4_fsblk_t n_blocks_count_retry = 0; unsigned long last_update_time = 0; - int err = 0, flexbg_size = 1 << sbi->s_log_groups_per_flex; + int err = 0; int meta_bg; + unsigned int flexbg_size = ext4_flex_bg_size(sbi);
/* See if the device is actually as big as what was requested */ bh = ext4_sb_bread(sb, n_blocks_count - 1, 0);
From: Baokun Li libaokun1@huawei.com
[ Upstream commit b099eb87de105cf07cad731ded6fb40b2675108b ]
In commit 967ac8af4475 ("ext4: fix potential integer overflow in alloc_flex_gd()"), an overflow check is added to alloc_flex_gd() to prevent the allocated memory from being smaller than expected due to the overflow. However, after kmalloc() is replaced with kmalloc_array() in commit 6da2ec56059c ("treewide: kmalloc() -> kmalloc_array()"), the kmalloc_array() function has an overflow check, so the above problem will not occur. Therefore, the extra check is removed.
Signed-off-by: Baokun Li libaokun1@huawei.com Reviewed-by: Jan Kara jack@suse.cz Link: https://lore.kernel.org/r/20231023013057.2117948-3-libaokun1@huawei.com Signed-off-by: Theodore Ts'o tytso@mit.edu Signed-off-by: Sasha Levin sashal@kernel.org --- fs/ext4/resize.c | 3 --- 1 file changed, 3 deletions(-)
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c index 96d278688fd7..66ce3d118203 100644 --- a/fs/ext4/resize.c +++ b/fs/ext4/resize.c @@ -245,10 +245,7 @@ static struct ext4_new_flex_group_data *alloc_flex_gd(unsigned int flexbg_size) if (flex_gd == NULL) goto out3;
- if (flexbg_size >= UINT_MAX / sizeof(struct ext4_new_group_data)) - goto out2; flex_gd->count = flexbg_size; - flex_gd->groups = kmalloc_array(flexbg_size, sizeof(struct ext4_new_group_data), GFP_NOFS);
From: Baokun Li libaokun1@huawei.com
[ Upstream commit 5d1935ac02ca5aee364a449a35e2977ea84509b0 ]
When we online resize an ext4 filesystem with a oversized flexbg_size,
mkfs.ext4 -F -G 67108864 $dev -b 4096 100M mount $dev $dir resize2fs $dev 16G
the following WARN_ON is triggered: ================================================================== WARNING: CPU: 0 PID: 427 at mm/page_alloc.c:4402 __alloc_pages+0x411/0x550 Modules linked in: sg(E) CPU: 0 PID: 427 Comm: resize2fs Tainted: G E 6.6.0-rc5+ #314 RIP: 0010:__alloc_pages+0x411/0x550 Call Trace: <TASK> __kmalloc_large_node+0xa2/0x200 __kmalloc+0x16e/0x290 ext4_resize_fs+0x481/0xd80 __ext4_ioctl+0x1616/0x1d90 ext4_ioctl+0x12/0x20 __x64_sys_ioctl+0xf0/0x150 do_syscall_64+0x3b/0x90 ==================================================================
This is because flexbg_size is too large and the size of the new_group_data array to be allocated exceeds MAX_ORDER. Currently, the minimum value of MAX_ORDER is 8, the minimum value of PAGE_SIZE is 4096, the corresponding maximum number of groups that can be allocated is:
(PAGE_SIZE << MAX_ORDER) / sizeof(struct ext4_new_group_data) ≈ 21845
And the value that is down-aligned to the power of 2 is 16384. Therefore, this value is defined as MAX_RESIZE_BG, and the number of groups added each time does not exceed this value during resizing, and is added multiple times to complete the online resizing. The difference is that the metadata in a flex_bg may be more dispersed.
Signed-off-by: Baokun Li libaokun1@huawei.com Reviewed-by: Jan Kara jack@suse.cz Link: https://lore.kernel.org/r/20231023013057.2117948-4-libaokun1@huawei.com Signed-off-by: Theodore Ts'o tytso@mit.edu Signed-off-by: Sasha Levin sashal@kernel.org --- fs/ext4/resize.c | 25 +++++++++++++++++-------- 1 file changed, 17 insertions(+), 8 deletions(-)
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c index 66ce3d118203..06e0eaf2ea4e 100644 --- a/fs/ext4/resize.c +++ b/fs/ext4/resize.c @@ -227,10 +227,17 @@ struct ext4_new_flex_group_data { in the flex group */ __u16 *bg_flags; /* block group flags of groups in @groups */ + ext4_group_t resize_bg; /* number of allocated + new_group_data */ ext4_group_t count; /* number of groups in @groups */ };
+/* + * Avoiding memory allocation failures due to too many groups added each time. + */ +#define MAX_RESIZE_BG 16384 + /* * alloc_flex_gd() allocates a ext4_new_flex_group_data with size of * @flexbg_size. @@ -245,14 +252,18 @@ static struct ext4_new_flex_group_data *alloc_flex_gd(unsigned int flexbg_size) if (flex_gd == NULL) goto out3;
- flex_gd->count = flexbg_size; - flex_gd->groups = kmalloc_array(flexbg_size, + if (unlikely(flexbg_size > MAX_RESIZE_BG)) + flex_gd->resize_bg = MAX_RESIZE_BG; + else + flex_gd->resize_bg = flexbg_size; + + flex_gd->groups = kmalloc_array(flex_gd->resize_bg, sizeof(struct ext4_new_group_data), GFP_NOFS); if (flex_gd->groups == NULL) goto out2;
- flex_gd->bg_flags = kmalloc_array(flexbg_size, sizeof(__u16), + flex_gd->bg_flags = kmalloc_array(flex_gd->resize_bg, sizeof(__u16), GFP_NOFS); if (flex_gd->bg_flags == NULL) goto out1; @@ -1559,8 +1570,7 @@ static int ext4_flex_group_add(struct super_block *sb,
static int ext4_setup_next_flex_gd(struct super_block *sb, struct ext4_new_flex_group_data *flex_gd, - ext4_fsblk_t n_blocks_count, - unsigned int flexbg_size) + ext4_fsblk_t n_blocks_count) { struct ext4_sb_info *sbi = EXT4_SB(sb); struct ext4_super_block *es = sbi->s_es; @@ -1584,7 +1594,7 @@ static int ext4_setup_next_flex_gd(struct super_block *sb, BUG_ON(last); ext4_get_group_no_and_offset(sb, n_blocks_count - 1, &n_group, &last);
- last_group = group | (flexbg_size - 1); + last_group = group | (flex_gd->resize_bg - 1); if (last_group > n_group) last_group = n_group;
@@ -2081,8 +2091,7 @@ int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count) /* Add flex groups. Note that a regular group is a * flex group with 1 group. */ - while (ext4_setup_next_flex_gd(sb, flex_gd, n_blocks_count, - flexbg_size)) { + while (ext4_setup_next_flex_gd(sb, flex_gd, n_blocks_count)) { if (jiffies - last_update_time > HZ * 10) { if (last_update_time) ext4_msg(sb, KERN_INFO,
linux-stable-mirror@lists.linaro.org
-
Sasha Levin