diff --git a/man2/arch_prctl.2 b/man2/arch_prctl.2
index 83da76db2..a6de7b9aa 100644
--- a/man2/arch_prctl.2
+++ b/man2/arch_prctl.2
@@ -118,7 +118,7 @@ As an optimization, if a 32-bit TLS base address is used,
 may use a real TLS entry as if
 .BR set_thread_area (2)
 had been called, instead of manipulating the segment base register directly.
-Memory in the first 2GB of address space can be allocated by using
+Memory in the first 2\ GB of address space can be allocated by using
 .BR mmap (2)
 with the
 .B MAP_32BIT
diff --git a/man2/ioctl_fideduperange.2 b/man2/ioctl_fideduperange.2
index 934f72c78..0ccd60a2c 100644
--- a/man2/ioctl_fideduperange.2
+++ b/man2/ioctl_fideduperange.2
@@ -116,7 +116,7 @@ and the struct
 array must not exceed the system page size.
 The maximum size of
 .IR src_length
-is filesystem dependent and is typically 16MiB.
+is filesystem dependent and is typically 16\ MiB.
 This limit will be enforced silently by the filesystem.
 By convention, the storage used by
 .IR src_fd
diff --git a/man2/ioctl_getfsmap.2 b/man2/ioctl_getfsmap.2
index f86f40926..aa1709596 100644
--- a/man2/ioctl_getfsmap.2
+++ b/man2/ioctl_getfsmap.2
@@ -132,8 +132,8 @@ to uniquely index any filesystem mapping record.
 Classic non-sharing filesystems might be able to identify any record with only
 .RI "(" "device" ", " "physical" ", " "flags" ")."
 For example, if the low key is set to (8:0, 36864, 0, 0, 0), the filesystem will
-only return records for extents starting at or above 36KiB on disk.
-If the high key is set to (8:0, 1048576, 0, 0, 0), only records below 1MiB will
+only return records for extents starting at or above 36\ KiB on disk.
+If the high key is set to (8:0, 1048576, 0, 0, 0), only records below 1\ MiB will
 be returned.
 The format of
 .I fmr_device
diff --git a/man2/kexec_load.2 b/man2/kexec_load.2
index 467caa1e4..5f9ad70f5 100644
--- a/man2/kexec_load.2
+++ b/man2/kexec_load.2
@@ -316,7 +316,7 @@ is empty (length zero).
 .I kernel_fd
 does not refer to an open file, or the kernel can't load this file.
 Currently, the file must be a bzImage and contain an x86 kernel that
-is loadable above 4GiB in memory (see the kernel source file
+is loadable above 4\ GiB in memory (see the kernel source file
 .IR Documentation/x86/boot.txt ).
 .TP
 .B ENOMEM
diff --git a/man2/madvise.2 b/man2/madvise.2
index 80cb63b40..dfb31b63d 100644
--- a/man2/madvise.2
+++ b/man2/madvise.2
@@ -336,8 +336,8 @@ naturally aligned to the huge page size (see
 This feature is primarily aimed at applications that use large mappings of
 data and access large regions of that memory at a time (e.g., virtualization
 systems such as QEMU).
-It can very easily waste memory (e.g., a 2MB mapping that only ever accesses
-1 byte will result in 2MB of wired memory instead of one 4KB page).
+It can very easily waste memory (e.g., a 2\ MB mapping that only ever accesses
+1 byte will result in 2\ MB of wired memory instead of one 4\ KB page).
 See the Linux kernel source file
 .I Documentation/vm/transhuge.txt
 for more details.
diff --git a/man2/mbind.2 b/man2/mbind.2
index fa2d99553..7b2b5eece 100644
--- a/man2/mbind.2
+++ b/man2/mbind.2
@@ -237,7 +237,7 @@ This optimizes for bandwidth instead of latency
 by spreading out pages and memory accesses to those pages across
 multiple nodes.
 To be effective the memory area should be fairly large,
-at least 1MB or bigger with a fairly uniform access pattern.
+at least 1\ MB or bigger with a fairly uniform access pattern.
 Accesses to a single page of the area will still be limited to
 the memory bandwidth of a single node.
 .TP
diff --git a/man2/mmap.2 b/man2/mmap.2
index 943c050cf..c95a64a6c 100644
--- a/man2/mmap.2
+++ b/man2/mmap.2
@@ -143,7 +143,7 @@ In addition, zero or more of the following values can be ORed in
 Put the mapping into the first 2 Gigabytes of the process address space.
 This flag is supported only on x86-64, for 64-bit programs.
 It was added to allow thread stacks to be allocated somewhere
-in the first 2GB of memory,
+in the first 2\ GB of memory,
 so as to improve context-switch performance on some early
 64-bit processors.
 .\" See http://lwn.net/Articles/294642 "Tangled up in threads", 19 Aug 08
diff --git a/man2/mmap2.2 b/man2/mmap2.2
index 7e99d781c..f3c654cd5 100644
--- a/man2/mmap2.2
+++ b/man2/mmap2.2
@@ -89,7 +89,7 @@ This system call does not exist on x86-64.
 On ia64, the unit for
 .I offset
 is actually the system page size, rather than 4096 bytes.
-.\" ia64 can have page sizes ranging from 4kB to 64kB.
+.\" ia64 can have page sizes ranging from 4 kB to 64 kB.
 .\" On cris, it looks like the unit might also be the page size,
 .\" which is 8192 bytes. -- mtk, June 2007
 .SH SEE ALSO
diff --git a/man2/mprotect.2 b/man2/mprotect.2
index 03d57a15e..755353466 100644
--- a/man2/mprotect.2
+++ b/man2/mprotect.2
@@ -197,7 +197,7 @@ was incorrectly produced for these cases.)
 Changing the protection of a memory region would result in the total number of
 mappings with distinct attributes (e.g., read versus read/write protection)
 exceeding the allowed maximum.
-.\" I.e., the number of VMAs would exceed the 64kB maximum
+.\" I.e., the number of VMAs would exceed the 64 kB maximum
 (For example, making the protection of a range
 .BR PROT_READ
 in the middle of a region currently protected as
diff --git a/man2/perf_event_open.2 b/man2/perf_event_open.2
index aea0a67f7..e373c16ea 100644
--- a/man2/perf_event_open.2
+++ b/man2/perf_event_open.2
@@ -1465,7 +1465,7 @@ struct perf_event_mmap_page {
     __u16 time_shift;
     __u32 time_mult;
     __u64 time_offset;
-    __u64 __reserved[120];   /* Pad to 1k */
+    __u64 __reserved[120];   /* Pad to 1 k */
     __u64 data_head;         /* head in the data section */
     __u64 data_tail;         /* user-space written tail */
     __u64 data_offset;       /* where the buffer starts */
diff --git a/man2/process_vm_readv.2 b/man2/process_vm_readv.2
index 3dbc42c6c..89a176d41 100644
--- a/man2/process_vm_readv.2
+++ b/man2/process_vm_readv.2
@@ -198,7 +198,7 @@ elements points to an invalid memory region in the remote process.
 No further reads/writes will be attempted beyond that point.
 Keep this in mind when attempting to read data of unknown length
 (such as C strings that are null-terminated) from a remote process,
-by avoiding spanning memory pages (typically 4KiB) in a single remote
+by avoiding spanning memory pages (typically 4\ KiB) in a single remote
 .I iovec
 element.
 (Instead, split the remote read into two
diff --git a/man2/set_mempolicy.2 b/man2/set_mempolicy.2
index c4a68f36e..8ee54d836 100644
--- a/man2/set_mempolicy.2
+++ b/man2/set_mempolicy.2
@@ -198,7 +198,7 @@ the memory bandwidth of a single node.
 .\" NOTE:  the following sentence doesn't make sense in the context
 .\" of set_mempolicy() -- no memory area specified.
 .\" To be effective the memory area should be fairly large,
-.\" at least 1MB or bigger.
+.\" at least 1 MB or bigger.
 .TP
 .B MPOL_PREFERRED
 This mode sets the preferred node for allocation.
diff --git a/man2/shmget.2 b/man2/shmget.2
index 023e1899a..c2c6d9c85 100644
--- a/man2/shmget.2
+++ b/man2/shmget.2
@@ -362,7 +362,7 @@ for a discussion of why this default value (rather than
 is used.
 .IP
 From Linux 2.2 up to Linux 3.15, the default value of
-this limit was 0x2000000 (32MB).
+this limit was 0x2000000 (32\ MB).
 .IP
 Because it is not possible to map just part of a shared memory segment,
 the amount of virtual memory places another limit on the maximum size of a
diff --git a/man2/subpage_prot.2 b/man2/subpage_prot.2
index a60835818..0f0c9497e 100644
--- a/man2/subpage_prot.2
+++ b/man2/subpage_prot.2
@@ -42,8 +42,8 @@ There is no glibc wrapper for this system call; see NOTES.
 The PowerPC-specific
 .BR subpage_prot ()
 system call provides the facility to control the access
-permissions on individual 4kB subpages on systems configured with
-a page size of 64kB.
+permissions on individual 4\ kB subpages on systems configured with
+a page size of 64\ kB.
 .PP
 The protection map is applied to the memory pages in the region starting at
 .I addr
@@ -54,9 +54,9 @@ Both of these arguments must be aligned to a 64-kB boundary.
 .PP
 The protection map is specified in the buffer pointed to by
 .IR map .
-The map has 2 bits per 4kB subpage;
-thus each 32-bit word specifies the protections of 16 4kB subpages
-inside a 64kB page
+The map has 2 bits per 4\ kB subpage;
+thus each 32-bit word specifies the protections of 16 4\ kB subpages
+inside a 64\ kB page
 (so, the number of 32-bit words pointed to by
 .I map
 should equate to the number of 64-kB pages specified by
@@ -120,7 +120,7 @@ Implicit in this is that the regions of the address space that are
 protected are switched to use 4-kB hardware pages rather than 64-kB
 hardware pages (on machines with hardware 64-kB page support).
 .\" In the initial implementation, it was the case that:
-.\"     In fact the whole process is switched to use 4k hardware pages when the
+.\"     In fact the whole process is switched to use 4 kB hardware pages when the
 .\"     subpage_prot system call is used, but this could be improved in future
 .\"     to switch only the affected segments.
 .\" But Paul Mackerass says (Oct 2010): I'm pretty sure we now only switch
diff --git a/man3/pthread_attr_init.3 b/man3/pthread_attr_init.3
index 8793a149a..23155154f 100644
--- a/man3/pthread_attr_init.3
+++ b/man3/pthread_attr_init.3
@@ -121,7 +121,7 @@ we see the following:
 .in +4n
 .nf
 .\" Results from glibc 2.8, SUSE 11.0; Oct 2008
-.RB "$" " ulimit \-s" "       # No stack limit ==> default stack size is 2MB"
+.RB "$" " ulimit \-s" "       # No stack limit ==> default stack size is 2 MB"
 unlimited
 .RB "$" " ./a.out"
 Thread attributes:
diff --git a/man3/pthread_create.3 b/man3/pthread_create.3
index ec2b87e02..5cc7f4da7 100644
--- a/man3/pthread_create.3
+++ b/man3/pthread_create.3
@@ -243,7 +243,7 @@ Joined with thread 3; returned value was SERVUS
 .fi
 .in
 .PP
-In the next run, the program explicitly sets a stack size of 1MB (using
+In the next run, the program explicitly sets a stack size of 1\ MB (using
 .BR pthread_attr_setstacksize (3))
 for the created threads:
 .PP
diff --git a/man3/pthread_getattr_np.3 b/man3/pthread_getattr_np.3
index e36dd664e..b8b79e09d 100644
--- a/man3/pthread_getattr_np.3
+++ b/man3/pthread_getattr_np.3
@@ -128,7 +128,7 @@ a thread is created using default attributes:
 .PP
 .in +4n
 .nf
-.RB "$" " ulimit \-s" "      # No stack limit ==> default stack size is 2MB"
+.RB "$" " ulimit \-s" "      # No stack limit ==> default stack size is 2 MB"
 unlimited
 .RB "$" " ./a.out"
 Attributes of created thread:
diff --git a/man3/regex.3 b/man3/regex.3
index 9ca0d79e4..ea536dd0f 100644
--- a/man3/regex.3
+++ b/man3/regex.3
@@ -293,7 +293,7 @@ Invalid use of the range operator; for example, the ending point of the range
 occurs prior to the starting point.
 .TP
 .B REG_ESIZE
-Compiled regular expression requires a pattern buffer larger than 64Kb.
+Compiled regular expression requires a pattern buffer larger than 64\ kB.
 This is not defined by POSIX.2.
 .TP
 .B REG_ESPACE
diff --git a/man4/wavelan.4 b/man4/wavelan.4
index 26ab51c51..355e12e84 100644
--- a/man4/wavelan.4
+++ b/man4/wavelan.4
@@ -73,7 +73,7 @@ or disable it
 As the NWID is stored in the card Permanent Storage Area, it will be
 reuse at any further invocation of the driver.
 .SS Frequency & channels
-For the 2.4GHz 2.00 Hardware, you are able to set the frequency by
+For the 2.4\ GHz 2.00 Hardware, you are able to set the frequency by
 specifying one of the 10 defined channels
 .RI ( 2.412,
 .I 2.422, 2.425, 2.4305, 2.432, 2.442, 2.452, 2.460, 2.462
diff --git a/man5/filesystems.5 b/man5/filesystems.5
index c692137d7..91eb77b6c 100644
--- a/man5/filesystems.5
+++ b/man5/filesystems.5
@@ -124,7 +124,7 @@ that was integrated into Linux in kernel 2.4.24.
 .B minix
 is the filesystem used in the Minix operating system, the first to run
 under Linux.
-It has a number of shortcomings, including a 64MB partition size
+It has a number of shortcomings, including a 64\ MB partition size
 limit, short filenames, and a single timestamp.
 It remains useful for floppies and RAM disks.
 .TP
diff --git a/man5/proc.5 b/man5/proc.5
index 2783db8c9..a64045710 100644
--- a/man5/proc.5
+++ b/man5/proc.5
@@ -2868,7 +2868,7 @@ binary, GDB can be used to
 examine the current state of any kernel data structures.
 .IP
 The total length of the file is the size of physical memory (RAM) plus
-4KB.
+4\ KiB.
 .TP
 .IR /proc/keys " (since Linux 2.6.10)"
 See
diff --git a/man7/bootparam.7 b/man7/bootparam.7
index 07a39233f..d40bda8ca 100644
--- a/man7/bootparam.7
+++ b/man7/bootparam.7
@@ -455,12 +455,12 @@ by using the following:
 The first two numbers are specified in units of kB.
 The default
 .I buf_size
-is 32kB, and the maximum size that can be specified is a
-ridiculous 16384kB.
+is 32k\ B, and the maximum size that can be specified is a
+ridiculous 16384\ kB.
 The
 .I write_threshold
 is the value at which the buffer is committed to tape, with a
-default value of 30kB.
+default value of 30\ kB.
 The maximum number of buffers varies
 with the number of drives detected, and has a default of two.
 An example usage would be:
diff --git a/man7/raw.7 b/man7/raw.7
index ce1780393..1daa5ebf9 100644
--- a/man7/raw.7
+++ b/man7/raw.7
@@ -164,7 +164,7 @@ Packet too big.
 Either Path MTU Discovery is enabled (the
 .B IP_MTU_DISCOVER
 socket flag) or the packet size exceeds the maximum allowed IPv4
-packet size of 64KB.
+packet size of 64\ kB.
 .TP
 .B EOPNOTSUPP
 Invalid flag has been passed to a socket call (like
diff --git a/man7/tcp.7 b/man7/tcp.7
index 36994b1f1..415c7690d 100644
--- a/man7/tcp.7
+++ b/man7/tcp.7
@@ -142,7 +142,7 @@ extensions.
 These include Protection Against Wrapped
 Sequence Numbers (PAWS), Window Scaling and Timestamps.
 Window scaling allows the use
-of large (> 64K) TCP windows in order to support links with high
+of large (> 64\ kB) TCP windows in order to support links with high
 latency or bandwidth.
 To make use of them, the send and receive buffer sizes must be increased.
 They can be set globally with the
@@ -576,7 +576,7 @@ the orphaned connection is reset and a warning is printed.
 This limit exists only to prevent simple denial-of-service attacks.
 Lowering this limit is not recommended.
 Network conditions might require you to increase the number of
-orphans allowed, but note that each orphan can eat up to ~64K
+orphans allowed, but note that each orphan can eat up to ~64\ kB
 of unswappable memory.
 The default initial value is set equal to the kernel parameter NR_FILE.
 This initial default is adjusted depending on the memory in the system.
@@ -589,8 +589,8 @@ If this number is exceeded, the kernel will begin
 dropping requests.
 The default value of 256 is increased to
 1024 when the memory present in the system is adequate or
-greater (>= 128Mb), and reduced to 128 for those systems with
-very low memory (<= 32Mb).
+greater (>= 128\ MB), and reduced to 128 for those systems with
+very low memory (<= 32\ MB).
 .IP
 Prior to Linux 2.6.20,
 .\" commit 72a3effaf633bcae9034b7e176bdbd78d64a71db
@@ -740,7 +740,7 @@ of these values, depending on memory available in the system.
 .I min
 minimum size of the receive buffer used by each TCP socket.
 The default value is the system page size.
-(On Linux 2.4, the default value is 4K, lowered to
+(On Linux 2.4, the default value is 4\ kB, lowered to
 .B PAGE_SIZE
 bytes in low-memory systems.)
 This value
@@ -775,7 +775,7 @@ This is not used to limit the size of the receive buffer declared using
 on a socket.
 The default value is calculated using the formula
 .IP
-    max(87380, min(4MB, \fItcp_mem\fP[1]*PAGE_SIZE/128))
+    max(87380, min(4\ MB, \fItcp_mem\fP[1]*PAGE_SIZE/128))
 .IP
 (On Linux 2.4, the default is 87380*2 bytes,
 lowered to 87380 in low-memory systems).
@@ -912,9 +912,9 @@ TCP Reno in wired networks and throughput over wireless links.
 .\" Since 2.1.36
 Enable RFC\ 1323 TCP window scaling.
 This feature allows the use of a large window
-(> 64K) on a TCP connection, should the other end support it.
+(> 64\ kB) on a TCP connection, should the other end support it.
 Normally, the 16 bit window length field in the TCP header
-limits the window size to less than 64K bytes.
+limits the window size to less than 64\ kB.
 If larger windows are desired, applications can increase the size of
 their socket buffers and the window scaling option will be employed.
 If
@@ -934,7 +934,7 @@ depending on memory available.
 .I min
 Minimum size of the send buffer used by each TCP socket.
 The default value is the system page size.
-(On Linux 2.4, the default value is 4K bytes.)
+(On Linux 2.4, the default value is 4\ kB.)
 This value is used to ensure that in memory pressure mode,
 allocations below this size will still succeed.
 This is not used to bound the size of the send buffer declared using
@@ -947,7 +947,7 @@ This value overwrites the initial default buffer size from
 the generic global
 .I /proc/sys/net/core/wmem_default
 defined for all protocols.
-The default value is 16K bytes.
+The default value is 16\ kB.
 .\" True in Linux 2.4 and 2.6
 If larger send buffer sizes are desired, this value
 should be increased (to affect all sockets).
@@ -964,10 +964,10 @@ This is not used to limit the size of the send buffer declared using
 on a socket.
 The default value is calculated using the formula
 .IP
-    max(65536, min(4MB, \fItcp_mem\fP[1]*PAGE_SIZE/128))
+    max(65536, min(4\ MB, \fItcp_mem\fP[1]*PAGE_SIZE/128))
 .IP
-(On Linux 2.4, the default value is 128K bytes,
-lowered 64K depending on low-memory systems.)
+(On Linux 2.4, the default value is 128\ kB,
+lowered 64\ kB depending on low-memory systems.)
 .RE
 .TP
 .IR tcp_workaround_signed_windows " (Boolean; default: disabled; since Linux 2.6.26)"
diff --git a/man7/units.7 b/man7/units.7
index 6e94d1fbe..eb003fffa 100644
--- a/man7/units.7
+++ b/man7/units.7
@@ -105,16 +105,16 @@ sloppy use of the prefixes "kilo" and "mega" started to become
 regarded as the "real true meaning" when computers were involved.
 But then disk technology changed, and disk sizes became arbitrary numbers.
 After a period of uncertainty all disk manufacturers settled on the
-standard, namely k=1000, M=1000k, G=1000M.
+standard, namely k=1000, M=1000\ k, G=1000\ M.
 .PP
-The situation was messy: in the 14k4 modems, k=1000; in the 1.44MB
+The situation was messy: in the 14k4 modems, k=1000; in the 1.44\ MB
 .\" also common: 14.4k modem
 diskettes, M=1024000; and so on.
 In 1998 the IEC approved the standard
 that defines the binary prefixes given above, enabling people
 to be precise and unambiguous.
 .PP
-Thus, today, MB = 1000000B and MiB = 1048576B.
+Thus, today, MB = 1000000\ B and MiB = 1048576\ B.
 .PP
 In the free software world programs are slowly
 being changed to conform.
diff --git a/man8/ld.so.8 b/man8/ld.so.8
index 66137b780..04d9b3fe8 100644
--- a/man8/ld.so.8
+++ b/man8/ld.so.8
@@ -620,14 +620,14 @@ If set to a nonempty string, warn about unresolved symbols.
 .BR LD_PREFER_MAP_32BIT_EXEC " (x86-64 only; since glibc 2.23)"
 According to the Intel Silvermont software optimization guide, for 64-bit
 applications, branch prediction performance can be negatively impacted
-when the target of a branch is more than 4GB away from the branch.
+when the target of a branch is more than 4\ GB away from the branch.
 If this environment variable is set (to any value),
 the dynamic linker
 will first try to map executable pages using the
 .BR mmap (2)
 .BR MAP_32BIT
 flag, and fall back to mapping without that flag if that attempt fails.
-NB: MAP_32BIT will map to the low 2GB (not 4GB) of the address space.
+NB: MAP_32BIT will map to the low 2\ GB (not 4\ GB) of the address space.
 .IP
 Because
 .B MAP_32BIT