On 3/20/25 6:55 AM, Jens Axboe wrote:
> On 3/19/25 5:51 PM, Greg Kroah-Hartman wrote:
>> On Wed, Mar 19, 2025 at 10:37:20AM -0600, Jens Axboe wrote:
>>> On 3/19/25 10:33 AM, Naresh Kamboju wrote:
>>>> On Wed, 19 Mar 2025 at 20:09, Greg Kroah-Hartman
>>>> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>>>>
>>>>> This is the start of the stable review cycle for the 6.6.84 release.
>>>>> There are 166 patches in this series, all will be posted as a response
>>>>> to this one. If anyone has any issues with these being applied, please
>>>>> let me know.
>>>>>
>>>>> Responses should be made by Fri, 21 Mar 2025 14:29:55 +0000.
>>>>> Anything received after that time might be too late.
>>>>>
>>>>> The whole patch series can be found in one patch at:
>>>>> https://www.kernel.org/pub/linux/kernel/v6.x/stable-review/patch-6.6.84-rc1.gz
>>>>> or in the git tree and branch at:
>>>>> git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git linux-6.6.y
>>>>> and the diffstat can be found below.
>>>>>
>>>>> thanks,
>>>>>
>>>>> greg k-h
>>>>
>>>> Regressions on mips the rt305x_defconfig builds failed with gcc-12
>>>> the stable-rc v6.6.83-167-gd16a828e7b09
>>>>
>>>> First seen on the v6.6.83-167-gd16a828e7b09
>>>> Good: v6.6.83
>>>> Bad: v6.6.83-167-gd16a828e7b09
>>>>
>>>> * mips, build
>>>> - gcc-12-rt305x_defconfig
>>>>
>>>> Regression Analysis:
>>>> - New regression? Yes
>>>> - Reproducibility? Yes
>>>>
>>>> Build regression: mips implicit declaration of function 'vunmap'
>>>> Reported-by: Linux Kernel Functional Testing <lkft@xxxxxxxxxx>
>>>
>>> Ah that's my fault, forgot to include the backport of:
>>>
>>> commit 62346c6cb28b043f2a6e95337d9081ec0b37b5f5
>>> Author: Jens Axboe <axboe@xxxxxxxxx>
>>> Date: Sat Mar 16 07:21:43 2024 -0600
>>>
>>> mm: add nommu variant of vm_insert_pages()
>>>
>>> for 6.1-stable and 6.6-stable. Greg, can you just cherry pick that one?
>>> It'll pick cleanly into both, should go before the io_uring mmap series
>>> obviously.
>>>
>>> Sorry about that! I did have it in my local trees, but for some reason
>>> forgot to include it in the sent in series.
>>
>> Wait, this is already in the 6.6.y and 6.1.y queues, so this can't be
>> the fix. Was there a fixup for that commit somewhere else that I'm
>> missing?
>
> Huh indeed, guess I didn't mess up in the first place. What is going on
> here indeed... Is that mips config NOMMU yet doesn't link in mm/nommu.o?
>
> Checking, and no, it definitely has MMU=y in the config. Guess I
> should've read the initial report more closely, it's simply missing the
> vunmap definition. Adding linux/vmalloc.h to io_uring/io_uring.c should
> fix it.
>
> How do we want to deal with this?
Either fold in the hunk from the previous email, or replace patch 2 in
the series I sent with this one instead, which adds the vmalloc.h
include. That should sort out the issue without needing add-on patches
that don't exist upstream.
--
Jens Axboe
From 063cda8806cf277c56e4bfc7163e5faa355e177d Mon Sep 17 00:00:00 2001
From: Jens Axboe <axboe@xxxxxxxxx>
Date: Wed, 13 Mar 2024 09:56:14 -0600
Subject: [PATCH 02/10] io_uring: get rid of remap_pfn_range() for mapping
rings/sqes
Commit 3ab1db3c6039e02a9deb9d5091d28d559917a645 upstream.
Rather than use remap_pfn_range() for this and manually free later,
switch to using vm_insert_pages() and have it Just Work.
If possible, allocate a single compound page that covers the range that
is needed. If that works, then we can just use page_address() on that
page. If we fail to get a compound page, allocate single pages and use
vmap() to map them into the kernel virtual address space.
This just covers the rings/sqes, the other remaining user of the mmap
remap_pfn_range() user will be converted separately. Once that is done,
we can kill the old alloc/free code.
Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
---
io_uring/io_uring.c | 140 +++++++++++++++++++++++++++++++++++++++++---
io_uring/io_uring.h | 2 +
2 files changed, 134 insertions(+), 8 deletions(-)
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 5776440f584c..b506064ca29d 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -71,6 +71,7 @@
#include <linux/io_uring.h>
#include <linux/audit.h>
#include <linux/security.h>
+#include <linux/vmalloc.h>
#include <asm/shmparam.h>
#define CREATE_TRACE_POINTS
@@ -2683,6 +2684,36 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
return READ_ONCE(rings->cq.head) == READ_ONCE(rings->cq.tail) ? ret : 0;
}
+static void io_pages_unmap(void *ptr, struct page ***pages,
+ unsigned short *npages)
+{
+ bool do_vunmap = false;
+
+ if (!ptr)
+ return;
+
+ if (*npages) {
+ struct page **to_free = *pages;
+ int i;
+
+ /*
+ * Only did vmap for the non-compound multiple page case.
+ * For the compound page, we just need to put the head.
+ */
+ if (PageCompound(to_free[0]))
+ *npages = 1;
+ else if (*npages > 1)
+ do_vunmap = true;
+ for (i = 0; i < *npages; i++)
+ put_page(to_free[i]);
+ }
+ if (do_vunmap)
+ vunmap(ptr);
+ kvfree(*pages);
+ *pages = NULL;
+ *npages = 0;
+}
+
void io_mem_free(void *ptr)
{
if (!ptr)
@@ -2787,8 +2818,8 @@ static void *io_sqes_map(struct io_ring_ctx *ctx, unsigned long uaddr,
static void io_rings_free(struct io_ring_ctx *ctx)
{
if (!(ctx->flags & IORING_SETUP_NO_MMAP)) {
- io_mem_free(ctx->rings);
- io_mem_free(ctx->sq_sqes);
+ io_pages_unmap(ctx->rings, &ctx->ring_pages, &ctx->n_ring_pages);
+ io_pages_unmap(ctx->sq_sqes, &ctx->sqe_pages, &ctx->n_sqe_pages);
} else {
io_pages_free(&ctx->ring_pages, ctx->n_ring_pages);
ctx->n_ring_pages = 0;
@@ -2800,6 +2831,80 @@ static void io_rings_free(struct io_ring_ctx *ctx)
ctx->sq_sqes = NULL;
}
+static void *io_mem_alloc_compound(struct page **pages, int nr_pages,
+ size_t size, gfp_t gfp)
+{
+ struct page *page;
+ int i, order;
+
+ order = get_order(size);
+ if (order > 10)
+ return ERR_PTR(-ENOMEM);
+ else if (order)
+ gfp |= __GFP_COMP;
+
+ page = alloc_pages(gfp, order);
+ if (!page)
+ return ERR_PTR(-ENOMEM);
+
+ for (i = 0; i < nr_pages; i++)
+ pages[i] = page + i;
+
+ return page_address(page);
+}
+
+static void *io_mem_alloc_single(struct page **pages, int nr_pages, size_t size,
+ gfp_t gfp)
+{
+ void *ret;
+ int i;
+
+ for (i = 0; i < nr_pages; i++) {
+ pages[i] = alloc_page(gfp);
+ if (!pages[i])
+ goto err;
+ }
+
+ ret = vmap(pages, nr_pages, VM_MAP, PAGE_KERNEL);
+ if (ret)
+ return ret;
+err:
+ while (i--)
+ put_page(pages[i]);
+ return ERR_PTR(-ENOMEM);
+}
+
+static void *io_pages_map(struct page ***out_pages, unsigned short *npages,
+ size_t size)
+{
+ gfp_t gfp = GFP_KERNEL_ACCOUNT | __GFP_ZERO | __GFP_NOWARN;
+ struct page **pages;
+ int nr_pages;
+ void *ret;
+
+ nr_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
+ pages = kvmalloc_array(nr_pages, sizeof(struct page *), gfp);
+ if (!pages)
+ return ERR_PTR(-ENOMEM);
+
+ ret = io_mem_alloc_compound(pages, nr_pages, size, gfp);
+ if (!IS_ERR(ret))
+ goto done;
+
+ ret = io_mem_alloc_single(pages, nr_pages, size, gfp);
+ if (!IS_ERR(ret)) {
+done:
+ *out_pages = pages;
+ *npages = nr_pages;
+ return ret;
+ }
+
+ kvfree(pages);
+ *out_pages = NULL;
+ *npages = 0;
+ return ret;
+}
+
void *io_mem_alloc(size_t size)
{
gfp_t gfp = GFP_KERNEL_ACCOUNT | __GFP_ZERO | __GFP_NOWARN | __GFP_COMP;
@@ -3463,14 +3568,12 @@ static void *io_uring_validate_mmap_request(struct file *file,
/* Don't allow mmap if the ring was setup without it */
if (ctx->flags & IORING_SETUP_NO_MMAP)
return ERR_PTR(-EINVAL);
- ptr = ctx->rings;
- break;
+ return ctx->rings;
case IORING_OFF_SQES:
/* Don't allow mmap if the ring was setup without it */
if (ctx->flags & IORING_SETUP_NO_MMAP)
return ERR_PTR(-EINVAL);
- ptr = ctx->sq_sqes;
- break;
+ return ctx->sq_sqes;
case IORING_OFF_PBUF_RING: {
struct io_buffer_list *bl;
unsigned int bgid;
@@ -3494,11 +3597,22 @@ static void *io_uring_validate_mmap_request(struct file *file,
return ptr;
}
+int io_uring_mmap_pages(struct io_ring_ctx *ctx, struct vm_area_struct *vma,
+ struct page **pages, int npages)
+{
+ unsigned long nr_pages = npages;
+
+ vm_flags_set(vma, VM_DONTEXPAND);
+ return vm_insert_pages(vma, vma->vm_start, pages, &nr_pages);
+}
+
#ifdef CONFIG_MMU
static __cold int io_uring_mmap(struct file *file, struct vm_area_struct *vma)
{
+ struct io_ring_ctx *ctx = file->private_data;
size_t sz = vma->vm_end - vma->vm_start;
+ long offset = vma->vm_pgoff << PAGE_SHIFT;
unsigned long pfn;
void *ptr;
@@ -3506,6 +3620,16 @@ static __cold int io_uring_mmap(struct file *file, struct vm_area_struct *vma)
if (IS_ERR(ptr))
return PTR_ERR(ptr);
+ switch (offset & IORING_OFF_MMAP_MASK) {
+ case IORING_OFF_SQ_RING:
+ case IORING_OFF_CQ_RING:
+ return io_uring_mmap_pages(ctx, vma, ctx->ring_pages,
+ ctx->n_ring_pages);
+ case IORING_OFF_SQES:
+ return io_uring_mmap_pages(ctx, vma, ctx->sqe_pages,
+ ctx->n_sqe_pages);
+ }
+
pfn = virt_to_phys(ptr) >> PAGE_SHIFT;
return remap_pfn_range(vma, vma->vm_start, pfn, sz, vma->vm_page_prot);
}
@@ -3795,7 +3919,7 @@ static __cold int io_allocate_scq_urings(struct io_ring_ctx *ctx,
return -EOVERFLOW;
if (!(ctx->flags & IORING_SETUP_NO_MMAP))
- rings = io_mem_alloc(size);
+ rings = io_pages_map(&ctx->ring_pages, &ctx->n_ring_pages, size);
else
rings = io_rings_map(ctx, p->cq_off.user_addr, size);
@@ -3820,7 +3944,7 @@ static __cold int io_allocate_scq_urings(struct io_ring_ctx *ctx,
}
if (!(ctx->flags & IORING_SETUP_NO_MMAP))
- ptr = io_mem_alloc(size);
+ ptr = io_pages_map(&ctx->sqe_pages, &ctx->n_sqe_pages, size);
else
ptr = io_sqes_map(ctx, p->sq_off.user_addr, size);
diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h
index 57658d24a73e..ea0b8acabc71 100644
--- a/io_uring/io_uring.h
+++ b/io_uring/io_uring.h
@@ -55,6 +55,8 @@ bool io_fill_cqe_req_aux(struct io_kiocb *req, bool defer, s32 res, u32 cflags);
void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
struct page **io_pin_pages(unsigned long ubuf, unsigned long len, int *npages);
+int io_uring_mmap_pages(struct io_ring_ctx *ctx, struct vm_area_struct *vma,
+ struct page **pages, int npages);
struct file *io_file_get_normal(struct io_kiocb *req, int fd);
struct file *io_file_get_fixed(struct io_kiocb *req, int fd,
--
2.49.0
