Skip to contentSkip to conversation

mm: add copy_remote_vm_str

PATCH Superseded Jordan Rome 2 participants 6 messages Jan 26, 2025 – Jan 28, 2025

Subsystems (6)

v5 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

PATCH v5 1/3 mm: add copy_remote_vm_str

Similar to access_process_vm but specific to strings. Also chunks reads by page and utilizes strscpy for handling null termination.

---
 include/linux/mm.h |   3 ++
 mm/memory.c        | 118 +++++++++++++++++++++++++++++++++++++++++++++
 mm/nommu.c         |  67 +++++++++++++++++++++++++
 3 files changed, 188 insertions(+)
include/linux/mm.h +3 -0
index f02925447e59..f3a05b3eb2f2 100644
@@ -2485,6 +2485,9 @@ extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
 		void *buf, int len, unsigned int gup_flags);
 
+extern int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+		void *buf, int len, unsigned int gup_flags);
+
 long get_user_pages_remote(struct mm_struct *mm,
 			   unsigned long start, unsigned long nr_pages,
 			   unsigned int gup_flags, struct page **pages,
mm/memory.c +118 -0

124 lines not on this page · raw message

mm/nommu.c +67 -0

73 lines not on this page · raw message

--
2.43.5

#1 lore raw

PATCH v5 2/3 bpf: Add bpf_copy_from_user_task_str kfunc

This new kfunc will be able to copy a string from another process's/task's address space. This is similar to bpf_copy_from_user_str but accepts a struct task_struct* argument.

---
 kernel/bpf/helpers.c | 48 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)
kernel/bpf/helpers.c +48 -0
index f27ce162427a..a33f72a4c31f 100644
@@ -3082,6 +3082,53 @@ __bpf_kfunc void bpf_local_irq_restore(unsigned long *flags__irq_flag)
 	local_irq_restore(*flags__irq_flag);
 }
 
+/**
+ * bpf_copy_from_user_task_str() - Copy a string from an task's address space
+ * @dst:             Destination address, in kernel space.  This buffer must be
+ *                   at least @dst__sz bytes long.
+ * @dst__sz:         Maximum number of bytes to copy, includes the trailing NUL.
+ * @unsafe_ptr__ign: Source address in the task's address space.
+ * @tsk:             The task whose address space will be used
+ * @flags:           The only supported flag is BPF_F_PAD_ZEROS
+ *
+ * Copies a NUL terminated string from a task's address space to @dst__sz
+ * buffer. If user string is too long this will still ensure zero termination
+ * in the @dst__sz buffer unless buffer size is 0.
+ *
+ * If BPF_F_PAD_ZEROS flag is set, memset the tail of @dst__sz to 0 on success
+ * and memset all of @dst__sz on failure.
+ *
+ * Return: The number of copied bytes on success including the NUL terminator.
+ * A negative error code on failure.
+ */
+__bpf_kfunc int bpf_copy_from_user_task_str(void *dst,
+					    u32 dst__sz,
+					    const void __user *unsafe_ptr__ign,
+					    struct task_struct *tsk,
+					    u64 flags)
+{
+	int ret;
+
+	if (unlikely(flags & ~BPF_F_PAD_ZEROS))
+		return -EINVAL;
+
+	if (unlikely(!dst__sz))
+		return 0;
+
+	ret = copy_remote_vm_str(tsk, (unsigned long)unsafe_ptr__ign, dst, dst__sz, 0);
+
+	if (ret < 0) {
+		if (flags & BPF_F_PAD_ZEROS)
+			memset(dst, 0, dst__sz);
+		return ret;
+	}
+
+	if (flags & BPF_F_PAD_ZEROS)
+		memset(dst + ret, 0, dst__sz - ret);
+
+	return ret + 1;
+}
+
 __bpf_kfunc_end_defs();
 
 BTF_KFUNCS_START(generic_btf_ids)
@@ -3174,6 +3221,7 @@ BTF_ID_FLAGS(func, bpf_iter_bits_new, KF_ITER_NEW)
 BTF_ID_FLAGS(func, bpf_iter_bits_next, KF_ITER_NEXT | KF_RET_NULL)
 BTF_ID_FLAGS(func, bpf_iter_bits_destroy, KF_ITER_DESTROY)
 BTF_ID_FLAGS(func, bpf_copy_from_user_str, KF_SLEEPABLE)
+BTF_ID_FLAGS(func, bpf_copy_from_user_task_str, KF_SLEEPABLE)
 BTF_ID_FLAGS(func, bpf_get_kmem_cache)
 BTF_ID_FLAGS(func, bpf_iter_kmem_cache_new, KF_ITER_NEW | KF_SLEEPABLE)
 BTF_ID_FLAGS(func, bpf_iter_kmem_cache_next, KF_ITER_NEXT | KF_RET_NULL | KF_SLEEPABLE)

--
2.43.5

Acked-by on v5 2/3
#2 lore raw

Re: [bpf-next v5 2/3] bpf: Add bpf_copy_from_user_task_str kfunc

Quoted 11 lines · Jordan Rome

This new kfunc will be able to copy a string from another process's/task's address space. This is similar to bpf_copy_from_user_str but accepts a struct task_struct* argument.

---
 kernel/bpf/helpers.c | 48 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)

Please carry over received acks/reviewed-by's, unless you drastically changed something about approved patch:

Quoted 7 lines
kernel/bpf/helpers.c +0 -0
index f27ce162427a..a33f72a4c31f 100644
@@ -3082,6 +3082,53 @@ __bpf_kfunc void bpf_local_irq_restore(unsigned long *flags__irq_flag)
        local_irq_restore(*flags__irq_flag);
 }

[...]

#3 lore raw

PATCH v5 3/3 selftests/bpf: Add tests for bpf_copy_from_user_task_str

In reply to Jordan Rome

This adds tests for both the happy path and the error path (with and without the BPF_F_PAD_ZEROS flag).

---
 .../selftests/bpf/prog_tests/bpf_iter.c       |  68 +++++++++++
 .../selftests/bpf/prog_tests/read_vsyscall.c  |   1 +
 .../selftests/bpf/progs/bpf_iter_tasks.c      | 110 ++++++++++++++++++
 .../selftests/bpf/progs/read_vsyscall.c       |  11 +-
 4 files changed, 188 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +68 -0
index 6f1bfacd7375..add4a18c33bd 100644
@@ -323,19 +323,87 @@ static void test_task_pidfd(void)
 static void test_task_sleepable(void)
 {
 	struct bpf_iter_tasks *skel;
+	int pid, status, err, data_pipe[2], finish_pipe[2], c;
+	char *test_data = NULL;
+	char *test_data_long = NULL;
+	char *data[2];
+
+	if (!ASSERT_OK(pipe(data_pipe), "data_pipe") ||
+	    !ASSERT_OK(pipe(finish_pipe), "finish_pipe"))
+		return;
 
 	skel = bpf_iter_tasks__open_and_load();
 	if (!ASSERT_OK_PTR(skel, "bpf_iter_tasks__open_and_load"))
 		return;
 
+	pid = fork();
+	if (!ASSERT_GE(pid, 0, "fork"))
+		return;
+
+	if (pid == 0) {
+		/* child */
+		close(data_pipe[0]);
+		close(finish_pipe[1]);
+
+		test_data = malloc(sizeof(char) * 10);
+		strncpy(test_data, "test_data", 10);
+		test_data[9] = '\0';
+
+		test_data_long = malloc(sizeof(char) * 5000);
+		for (int i = 0; i < 5000; ++i) {
+			if (i % 2 == 0)
+				test_data_long[i] = 'b';
+			else
+				test_data_long[i] = 'a';
+		}
+		test_data_long[4999] = '\0';
+
+		data[0] = test_data;
+		data[1] = test_data_long;
+
+		write(data_pipe[1], &data, sizeof(data));
+
+		/* keep child alive until after the test */
+		err = read(finish_pipe[0], &c, 1);
+		if (err != 1)
+			exit(-1);
+
+		close(data_pipe[1]);
+		close(finish_pipe[0]);
+		_exit(0);
+	}
+
+	/* parent */
+	close(data_pipe[1]);
+	close(finish_pipe[0]);
+
+	err = read(data_pipe[0], &data, sizeof(data));
+	ASSERT_EQ(err, sizeof(data), "read_check");
+
+	skel->bss->user_ptr = data[0];
+	skel->bss->user_ptr_long = data[1];
+	skel->bss->pid = pid;
+
 	do_dummy_read(skel->progs.dump_task_sleepable);
 
 	ASSERT_GT(skel->bss->num_expected_failure_copy_from_user_task, 0,
 		  "num_expected_failure_copy_from_user_task");
 	ASSERT_GT(skel->bss->num_success_copy_from_user_task, 0,
 		  "num_success_copy_from_user_task");
+	ASSERT_GT(skel->bss->num_expected_failure_copy_from_user_task_str, 0,
+		  "num_expected_failure_copy_from_user_task_str");
+	ASSERT_GT(skel->bss->num_success_copy_from_user_task_str, 0,
+		  "num_success_copy_from_user_task_str");
 
 	bpf_iter_tasks__destroy(skel);
+
+	write(finish_pipe[1], &c, 1);
+	err = waitpid(pid, &status, 0);
+	ASSERT_EQ(err, pid, "waitpid");
+	ASSERT_EQ(status, 0, "zero_child_exit");
+
+	close(data_pipe[0]);
+	close(finish_pipe[1]);
 }
 
 static void test_task_stack(void)
tools/testing/selftests/bpf/prog_tests/read_vsyscall.c +1 -0
index c7b9ba8b1d06..a8d1eaa67020 100644
@@ -24,6 +24,7 @@ struct read_ret_desc {
 	{ .name = "copy_from_user", .ret = -EFAULT },
 	{ .name = "copy_from_user_task", .ret = -EFAULT },
 	{ .name = "copy_from_user_str", .ret = -EFAULT },
+	{ .name = "copy_from_user_task_str", .ret = -EFAULT },
 };
 
 void test_read_vsyscall(void)
tools/testing/selftests/bpf/progs/bpf_iter_tasks.c +110 -0

137 lines not on this page · raw message

tools/testing/selftests/bpf/progs/read_vsyscall.c +9 -2
index 39ebef430059..395591374d4f 100644
@@ -8,14 +8,16 @@
 
 int target_pid = 0;
 void *user_ptr = 0;
-int read_ret[9];
+int read_ret[10];
 
 char _license[] SEC("license") = "GPL";
 
 /*
- * This is the only kfunc, the others are helpers
+ * These are the kfuncs, the others are helpers
  */
 int bpf_copy_from_user_str(void *dst, u32, const void *, u64) __weak __ksym;
+int bpf_copy_from_user_task_str(void *dst, u32, const void *,
+				struct task_struct *, u64) __weak __ksym;
 
 SEC("fentry/" SYS_PREFIX "sys_nanosleep")
 int do_probe_read(void *ctx)
@@ -47,6 +49,11 @@ int do_copy_from_user(void *ctx)
 	read_ret[7] = bpf_copy_from_user_task(buf, sizeof(buf), user_ptr,
 					      bpf_get_current_task_btf(), 0);
 	read_ret[8] = bpf_copy_from_user_str((char *)buf, sizeof(buf), user_ptr, 0);
+	read_ret[9] = bpf_copy_from_user_task_str((char *)buf,
+						  sizeof(buf),
+						  user_ptr,
+						  bpf_get_current_task_btf(),
+						  0);
 
 	return 0;
 }

--
2.43.5

#4 lore raw
In reply to Jordan Rome
Quoted 74 lines · Jordan Rome

Similar to access_process_vm but specific to strings. Also chunks reads by page and utilizes strscpy for handling null termination.

---
 include/linux/mm.h |   3 ++
 mm/memory.c        | 118 +++++++++++++++++++++++++++++++++++++++++++++
 mm/nommu.c         |  67 +++++++++++++++++++++++++
 3 files changed, 188 insertions(+)
include/linux/mm.h +3 -0
index f02925447e59..f3a05b3eb2f2 100644
@@ -2485,6 +2485,9 @@ extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
                void *buf, int len, unsigned int gup_flags);
 
+extern int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+               void *buf, int len, unsigned int gup_flags);
+
 long get_user_pages_remote(struct mm_struct *mm,
                           unsigned long start, unsigned long nr_pages,
                           unsigned int gup_flags, struct page **pages,
mm/memory.c +40 -0
index 398c031be9ba..e1ed5095b258 100644
@@ -6714,6 +6714,124 @@ int access_process_vm(struct task_struct *tsk, unsigned long addr,
 }
 EXPORT_SYMBOL_GPL(access_process_vm);
 
+/*
+ * Copy a string from another process's address space as given in mm.
+ * If there is any error return -EFAULT.
+ */
+static int __copy_remote_vm_str(struct mm_struct *mm, unsigned long addr,
+                             void *buf, int len, unsigned int gup_flags)
+{
+       void *old_buf = buf;
+       int err = 0;
+
+       if (mmap_read_lock_killable(mm))
+               return -EFAULT;
+
+       /* Untag the address before looking up the VMA */
+       addr = untagged_addr_remote(mm, addr);
+
+       /* Avoid triggering the temporary warning in __get_user_pages */
+       if (!vma_lookup(mm, addr)) {
+               err = -EFAULT;
+               goto out;
+       }
+
+       while (len) {
+               int bytes, offset, retval;
+               void *maddr;
+               struct page *page;
+               struct vm_area_struct *vma = NULL;
+
+               page = get_user_page_vma_remote(mm, addr, gup_flags, &vma);
+
+               if (IS_ERR(page)) {
+                       /*
+                        * Treat as a total failure for now until we decide how
+                        * to handle the CONFIG_HAVE_IOREMAP_PROT case and
+                        * stack expansion.
+                        */
+                       err = -EFAULT;
+                       goto out;
+               }
+

mm/memory.c int access_process_vm(struct task_struct *tsk, unsigned long addr, · line 6766

Quoted 10 lines
mm/memory.c +10 -0
+               bytes = len;
+               offset = addr & (PAGE_SIZE - 1);
+               if (bytes > PAGE_SIZE - offset)
+                       bytes = PAGE_SIZE - offset;
+
+               maddr = kmap_local_page(page);
+               retval = strscpy(buf, maddr + offset, bytes);
+               unmap_and_put_page(page, maddr);
+
+               if (retval > -1) {

nit: retval >= 0 is more conventional, -1 has no special meaning here (it's -EPERM, if anything), zero does

Quoted 14 lines
mm/memory.c +14 -0
+                       /* Found the end of the string */
+                       buf += retval;
+                       goto out;
+               }
+
+               retval = bytes - 1;
+               buf += retval;
+
+               if (bytes == len)
+                       goto out;
+
+               /*
+                * Because strscpy always NUL terminates we need to
+                * copy the last byte in the page if we are going to

mm/memory.c int access_process_vm(struct task_struct *tsk, unsigned long addr, · line 6790

Quoted 10 lines
mm/memory.c +10 -0
+                * load more pages
+                */
+               addr += retval;
+               len -= retval;
+               copy_from_user_page(vma,
+                               page,
+                               addr,
+                               buf,
+                               maddr + (PAGE_SIZE - 1),
+                               1);

just realized, you've already unmap_and_put_page(), and yet you are trying to access it here again. Seems like you'll need to delay that unmap_and_put

also, stylistical nit: you have tons of short arguments, keep them on smaller number of lines, it's taking way too much vertical space for what it is

Also, I was worried about non-zero terminated buf here. It still can happen if subsequent get_user_page_vma_remote() fails and we exit early, but we'll end up returning -EFAULT, so perhaps that's not a problem. On the other hand, it's trivial to do buf[1] = '\0'; to keep that buf always zero-terminated, so maybe I'd do that... And if we do buf[0] = '\0'; at the very beginning, we can document that copy_remote_vm_str() leaves valid zero-terminated buffer of whatever it managed to read before EFAULT, which isn't a bad property that basically comes for free, no?

pw-bot: cr

Quoted 15 lines
mm/memory.c +15 -0
+               len -= 1;
+               buf += 1;
+               addr += 1;
+       }
+
+out:
+       mmap_read_unlock(mm);
+       if (err)
+               return err;
+
+       return buf - old_buf;
+}
+
+/**
+ * copy_remote_vm_str - copy a string from another process's address space.

mm/memory.c int access_process_vm(struct task_struct *tsk, unsigned long addr, · line 6815

Quoted 10 lines
mm/memory.c +10 -0
+ * @tsk:       the task of the target address space
+ * @addr:      start address to read from
+ * @buf:       destination buffer
+ * @len:       number of bytes to copy
+ * @gup_flags: flags modifying lookup behaviour
+ *
+ * The caller must hold a reference on @mm.
+ *
+ * Return: number of bytes copied from @addr (source) to @buf (destination);
+ * not including the trailing NUL. On any error, return -EFAULT.

"On success, leaves a properly NUL-terminated buffer." (or if we do adjustments I suggested above we can even say "Even on error will leave properly NUL-terminated buffer with contents that was successfully read before -EFAULT", or something along those lines).

Quoted 38 lines
mm/memory.c +19 -0
+ */
+int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+               void *buf, int len, unsigned int gup_flags)
+{
+       struct mm_struct *mm;
+       int ret;
+
+       mm = get_task_mm(tsk);
+       if (!mm)
+               return -EFAULT;
+
+       ret = __copy_remote_vm_str(mm, addr, buf, len, gup_flags);
+
+       mmput(mm);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(copy_remote_vm_str);
+
 /*
  * Print the name of a VMA.
  */
mm/nommu.c +8 -0
index 9cb6e99215e2..ce24ea829c73 100644
@@ -1701,6 +1701,73 @@ int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, in
 }
 EXPORT_SYMBOL_GPL(access_process_vm);
 
+/*
+ * Copy a string from another process's address space as given in mm.
+ * If there is any error return -EFAULT.
+ */
+static int __copy_remote_vm_str(struct mm_struct *mm, unsigned long addr,
+                             void *buf, int len)
+{
+       int ret;

mm/nommu.c int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, in · line 1721

Quoted 10 lines
mm/nommu.c +10 -0
+       struct vm_area_struct *vma;
+
+       if (mmap_read_lock_killable(mm))
+               return -EFAULT;
+
+       /* the access must start within one of the target process's mappings */
+       vma = find_vma(mm, addr);
+       if (vma) {
+               /* don't overrun this mapping */
+               if (addr + len >= vma->vm_end)

Should we worry about overflows here? check_add_overflow() maybe?

Quoted 4 lines
mm/nommu.c +4 -0
+                       len = vma->vm_end - addr;
+
+               /* only read mappings where it is permitted */
+               if (vma->vm_flags & VM_MAYREAD) {

mm/nommu.c int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, in · line 1735

Quoted 10 lines
mm/nommu.c +10 -0
+                       ret = strscpy(buf, (char *)addr, len);
+                       if (ret < 0)
+                               ret = len - 1;
+               } else {
+                       ret = -EFAULT;
+               }
+       } else {
+               ret = -EFAULT;
+       }
+

nit: might be cleaner to have ret = -EFAULT; before if (vma) and only set ret on success/overflow?

Quoted 42 lines
mm/nommu.c +35 -1
+       mmap_read_unlock(mm);
+       return ret;
+}
+
+/**
+ * copy_remote_vm_str - copy a string from another process's address space.
+ * @tsk:       the task of the target address space
+ * @addr:      start address to read from
+ * @buf:       destination buffer
+ * @len:       number of bytes to copy
+ * @gup_flags: flags modifying lookup behaviour (unused)
+ *
+ * The caller must hold a reference on @mm.
+ *
+ * Return: number of bytes copied from @addr (source) to @buf (destination);
+ * not including the trailing NUL. On any error, return -EFAULT.
+ */
+int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+               void *buf, int len, unsigned int gup_flags)
+{
+       struct mm_struct *mm;
+       int ret;
+
+       mm = get_task_mm(tsk);
+       if (!mm)
+               return -EFAULT;
+
+       ret = __copy_remote_vm_str(mm, addr, buf, len);
+
+       mmput(mm);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(copy_remote_vm_str);
+
 /**
  * nommu_shrink_inode_mappings - Shrink the shared mappings on an inode
  * @inode: The inode to check
--

2.43.5

#5 lore raw
Quoted 81 lines · Andrii Nakryiko
Quoted 79 lines · Jordan Rome

Similar to access_process_vm but specific to strings. Also chunks reads by page and utilizes strscpy for handling null termination.

---
 include/linux/mm.h |   3 ++
 mm/memory.c        | 118 +++++++++++++++++++++++++++++++++++++++++++++
 mm/nommu.c         |  67 +++++++++++++++++++++++++
 3 files changed, 188 insertions(+)
include/linux/mm.h +3 -0
index f02925447e59..f3a05b3eb2f2 100644
@@ -2485,6 +2485,9 @@ extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
                void *buf, int len, unsigned int gup_flags);
 
+extern int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+               void *buf, int len, unsigned int gup_flags);
+
 long get_user_pages_remote(struct mm_struct *mm,
                           unsigned long start, unsigned long nr_pages,
                           unsigned int gup_flags, struct page **pages,
mm/memory.c +45 -0
index 398c031be9ba..e1ed5095b258 100644
@@ -6714,6 +6714,124 @@ int access_process_vm(struct task_struct *tsk, unsigned long addr,
 }
 EXPORT_SYMBOL_GPL(access_process_vm);
 
+/*
+ * Copy a string from another process's address space as given in mm.
+ * If there is any error return -EFAULT.
+ */
+static int __copy_remote_vm_str(struct mm_struct *mm, unsigned long addr,
+                             void *buf, int len, unsigned int gup_flags)
+{
+       void *old_buf = buf;
+       int err = 0;
+
+       if (mmap_read_lock_killable(mm))
+               return -EFAULT;
+
+       /* Untag the address before looking up the VMA */
+       addr = untagged_addr_remote(mm, addr);
+
+       /* Avoid triggering the temporary warning in __get_user_pages */
+       if (!vma_lookup(mm, addr)) {
+               err = -EFAULT;
+               goto out;
+       }
+
+       while (len) {
+               int bytes, offset, retval;
+               void *maddr;
+               struct page *page;
+               struct vm_area_struct *vma = NULL;
+
+               page = get_user_page_vma_remote(mm, addr, gup_flags, &vma);
+
+               if (IS_ERR(page)) {
+                       /*
+                        * Treat as a total failure for now until we decide how
+                        * to handle the CONFIG_HAVE_IOREMAP_PROT case and
+                        * stack expansion.
+                        */
+                       err = -EFAULT;
+                       goto out;
+               }
+
+               bytes = len;
+               offset = addr & (PAGE_SIZE - 1);
+               if (bytes > PAGE_SIZE - offset)
+                       bytes = PAGE_SIZE - offset;
+
Quoted 9 lines

mm/memory.c int access_process_vm(struct task_struct *tsk, unsigned long addr, · line 6766

Quoted 5 lines
mm/memory.c +5 -0
+               maddr = kmap_local_page(page);
+               retval = strscpy(buf, maddr + offset, bytes);
+               unmap_and_put_page(page, maddr);
+
+               if (retval > -1) {

nit: retval >= 0 is more conventional, -1 has no special meaning here (it's -EPERM, if anything), zero does

Ack.

Quoted 20 lines
Quoted 20 lines
mm/memory.c +20 -0
+                       /* Found the end of the string */
+                       buf += retval;
+                       goto out;
+               }
+
+               retval = bytes - 1;
+               buf += retval;
+
+               if (bytes == len)
+                       goto out;
+
+               /*
+                * Because strscpy always NUL terminates we need to
+                * copy the last byte in the page if we are going to
+                * load more pages
+                */
+               addr += retval;
+               len -= retval;
+               copy_from_user_page(vma,
+                               page,
Quoted 9 lines

mm/memory.c int access_process_vm(struct task_struct *tsk, unsigned long addr, · line 6790

Quoted 4 lines
mm/memory.c +4 -0
+                               addr,
+                               buf,
+                               maddr + (PAGE_SIZE - 1),
+                               1);

just realized, you've already unmap_and_put_page(), and yet you are trying to access it here again. Seems like you'll need to delay that unmap_and_put

Good catch on the unmap_and_put_page - will fix.

Quoted 4 lines

also, stylistical nit: you have tons of short arguments, keep them on smaller number of lines, it's taking way too much vertical space for what it is

Ack.

Quoted 9 lines

Also, I was worried about non-zero terminated buf here. It still can happen if subsequent get_user_page_vma_remote() fails and we exit early, but we'll end up returning -EFAULT, so perhaps that's not a problem. On the other hand, it's trivial to do buf[1] = '\0'; to keep that buf always zero-terminated, so maybe I'd do that... And if we do buf[0] = '\0'; at the very beginning, we can document that copy_remote_vm_str() leaves valid zero-terminated buffer of whatever it managed to read before EFAULT, which isn't a bad property that basically comes for free, no?

Sounds good. Will update to always nul terminate the buffer. It is a bit odd to me that users would read the buffer at all if an error is returned but, at the same time, this implementation is doing exactly that if strscopy returns E2BIG.

Quoted 25 lines

pw-bot: cr

Quoted 22 lines
mm/memory.c +22 -0
+               len -= 1;
+               buf += 1;
+               addr += 1;
+       }
+
+out:
+       mmap_read_unlock(mm);
+       if (err)
+               return err;
+
+       return buf - old_buf;
+}
+
+/**
+ * copy_remote_vm_str - copy a string from another process's address space.
+ * @tsk:       the task of the target address space
+ * @addr:      start address to read from
+ * @buf:       destination buffer
+ * @len:       number of bytes to copy
+ * @gup_flags: flags modifying lookup behaviour
+ *
+ * The caller must hold a reference on @mm.
Quoted 9 lines

mm/memory.c int access_process_vm(struct task_struct *tsk, unsigned long addr, · line 6815

Quoted 3 lines
mm/memory.c +3 -0
+ *
+ * Return: number of bytes copied from @addr (source) to @buf (destination);
+ * not including the trailing NUL. On any error, return -EFAULT.

"On success, leaves a properly NUL-terminated buffer." (or if we do adjustments I suggested above we can even say "Even on error will leave properly NUL-terminated buffer with contents that was successfully read before -EFAULT", or something along those lines).

Ack.

Quoted 42 lines
Quoted 42 lines
mm/memory.c +19 -0
+ */
+int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+               void *buf, int len, unsigned int gup_flags)
+{
+       struct mm_struct *mm;
+       int ret;
+
+       mm = get_task_mm(tsk);
+       if (!mm)
+               return -EFAULT;
+
+       ret = __copy_remote_vm_str(mm, addr, buf, len, gup_flags);
+
+       mmput(mm);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(copy_remote_vm_str);
+
 /*
  * Print the name of a VMA.
  */
mm/nommu.c +12 -0
index 9cb6e99215e2..ce24ea829c73 100644
@@ -1701,6 +1701,73 @@ int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, in
 }
 EXPORT_SYMBOL_GPL(access_process_vm);
 
+/*
+ * Copy a string from another process's address space as given in mm.
+ * If there is any error return -EFAULT.
+ */
+static int __copy_remote_vm_str(struct mm_struct *mm, unsigned long addr,
+                             void *buf, int len)
+{
+       int ret;
+       struct vm_area_struct *vma;
+
+       if (mmap_read_lock_killable(mm))
+               return -EFAULT;
Quoted 8 lines

mm/nommu.c int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, in · line 1721

Quoted 6 lines
mm/nommu.c +6 -0
+
+       /* the access must start within one of the target process's mappings */
+       vma = find_vma(mm, addr);
+       if (vma) {
+               /* don't overrun this mapping */
+               if (addr + len >= vma->vm_end)

Should we worry about overflows here? check_add_overflow() maybe?

Ack.

Quoted 10 lines
Quoted 10 lines
mm/nommu.c +9 -0
+                       len = vma->vm_end - addr;
+
+               /* only read mappings where it is permitted */
+               if (vma->vm_flags & VM_MAYREAD) {
+                       ret = strscpy(buf, (char *)addr, len);
+                       if (ret < 0)
+                               ret = len - 1;
+               } else {
+                       ret = -EFAULT;
Quoted 8 lines

mm/nommu.c int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, in · line 1735

Quoted 5 lines
mm/nommu.c +5 -0
+               }
+       } else {
+               ret = -EFAULT;
+       }
+

nit: might be cleaner to have ret = -EFAULT; before if (vma) and only set ret on success/overflow?

Ack.

Quoted 43 lines
Quoted 43 lines
mm/nommu.c +35 -1
+       mmap_read_unlock(mm);
+       return ret;
+}
+
+/**
+ * copy_remote_vm_str - copy a string from another process's address space.
+ * @tsk:       the task of the target address space
+ * @addr:      start address to read from
+ * @buf:       destination buffer
+ * @len:       number of bytes to copy
+ * @gup_flags: flags modifying lookup behaviour (unused)
+ *
+ * The caller must hold a reference on @mm.
+ *
+ * Return: number of bytes copied from @addr (source) to @buf (destination);
+ * not including the trailing NUL. On any error, return -EFAULT.
+ */
+int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
+               void *buf, int len, unsigned int gup_flags)
+{
+       struct mm_struct *mm;
+       int ret;
+
+       mm = get_task_mm(tsk);
+       if (!mm)
+               return -EFAULT;
+
+       ret = __copy_remote_vm_str(mm, addr, buf, len);
+
+       mmput(mm);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(copy_remote_vm_str);
+
 /**
  * nommu_shrink_inode_mappings - Shrink the shared mappings on an inode
  * @inode: The inode to check
--

2.43.5